coh_ocr_proj.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "custom mmocr.ipynb",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3",
      "language": "python"
    },
    "language_info": {
      "name": "python",
      "version": "3.8.5"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/balandongiv/mmocr_tutorial/blob/main/coh_ocr_proj.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# MMOCR Tutorial\n",
        "\n",
        "Welcome to MMOCR! This is the **unofficial** colab tutorial for using MMOCR. In this tutorial, you will learn how to\n",
        "\n",
        "- Transform annotation in coco-like json format to MMOCR compatible format.\n",
        "- Set custom configuration file\n",
        "- Integrate mmocr config template extracted from `base` with user custom setting\n",
        "- Train a text recognizer with a toy dataset. \n",
        "- Resume training from exisiting `training model`.\n",
        "- Load a pretrained model"
      ],
      "metadata": {
        "id": "U-Xyj4cQrEQ1"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Authors: Rodney Petrus Balandong balandongiv@gmail.com\n",
        "\n",
        "Part of noteebook content were extracted from the MMOCR official documentation"
      ],
      "metadata": {
        "id": "m_XABoECyCJA"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Install MMOCR"
      ],
      "metadata": {
        "id": "Sfvz1sywQ9_4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "When installing dependencies for mmocr, please ensure that all the dependency versions are compatible with each other. For instance, if CUDA 10.1 is installed, then the Pytorch version must be compatible with cu10.1. Please see [getting_started.md](docs/getting_started.md) for more details. "
      ],
      "metadata": {
        "id": "q3fZP1LspEUp"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "%cd .."
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/\n"
          ]
        }
      ],
      "metadata": {
        "id": "rB3qciTXpEUq",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1b5e0f8b-a202-4aef-b4ff-c5607a2f83c9"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Check PyTorch version"
      ],
      "metadata": {
        "id": "mSkZOdrMpEUr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip show torch"
      ],
      "metadata": {
        "id": "eNvnx5XGAn0i",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5e83d4f7-09fc-44e2-c57a-dfde79fa8d82"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Name: torch\n",
            "Version: 1.11.0+cu113\n",
            "Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration\n",
            "Home-page: https://pytorch.org/\n",
            "Author: PyTorch Team\n",
            "Author-email: packages@pytorch.org\n",
            "License: BSD-3\n",
            "Location: /usr/local/lib/python3.7/dist-packages\n",
            "Requires: typing-extensions\n",
            "Required-by: torchvision, torchtext, torchaudio, fastai\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Install Dependencies "
      ],
      "metadata": {
        "id": "Tw7u_baQpEUs"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Install mmcv-full thus we could use CUDA operators\n",
        "!pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.11.0/index.html\n",
        "\n",
        "# Install mmdetection\n",
        "!pip install mmdet\n",
        "\n",
        "# # Install mmocr\n",
        "!git clone https://github.com/open-mmlab/mmocr.git\n",
        "%cd mmocr\n",
        "!pip install -r requirements.txt\n",
        "!pip install -v -e .\n",
        "\n",
        "!pip install gdown"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Looking in links: https://download.openmmlab.com/mmcv/dist/cu113/torch1.11.0/index.html\n",
            "Collecting mmcv-full\n",
            "  Downloading https://download.openmmlab.com/mmcv/dist/cu113/torch1.11.0/mmcv_full-1.5.3-cp37-cp37m-manylinux1_x86_64.whl (38.0 MB)\n",
            "\u001b[K     |████████████████████████████████| 38.0 MB 38 kB/s \n",
            "\u001b[?25hCollecting yapf\n",
            "  Downloading yapf-0.32.0-py2.py3-none-any.whl (190 kB)\n",
            "\u001b[K     |████████████████████████████████| 190 kB 30.7 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pyyaml in /usr/local/lib/python3.7/dist-packages (from mmcv-full) (3.13)\n",
            "Requirement already satisfied: Pillow in /usr/local/lib/python3.7/dist-packages (from mmcv-full) (7.1.2)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.7/dist-packages (from mmcv-full) (21.3)\n",
            "Collecting addict\n",
            "  Downloading addict-2.4.0-py3-none-any.whl (3.8 kB)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from mmcv-full) (1.21.6)\n",
            "Requirement already satisfied: opencv-python>=3 in /usr/local/lib/python3.7/dist-packages (from mmcv-full) (4.1.2.30)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging->mmcv-full) (3.0.9)\n",
            "Installing collected packages: yapf, addict, mmcv-full\n",
            "Successfully installed addict-2.4.0 mmcv-full-1.5.3 yapf-0.32.0\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting mmdet\n",
            "  Downloading mmdet-2.25.0-py3-none-any.whl (1.4 MB)\n",
            "\u001b[K     |████████████████████████████████| 1.4 MB 26.6 MB/s \n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from mmdet) (1.15.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from mmdet) (1.21.6)\n",
            "Requirement already satisfied: pycocotools in /usr/local/lib/python3.7/dist-packages (from mmdet) (2.0.4)\n",
            "Collecting terminaltables\n",
            "  Downloading terminaltables-3.1.10-py2.py3-none-any.whl (15 kB)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from mmdet) (3.2.2)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmdet) (0.11.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmdet) (3.0.9)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmdet) (2.8.2)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmdet) (1.4.3)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->mmdet) (4.1.1)\n",
            "Installing collected packages: terminaltables, mmdet\n",
            "Successfully installed mmdet-2.25.0 terminaltables-3.1.10\n",
            "Cloning into 'mmocr'...\n",
            "remote: Enumerating objects: 6743, done.\u001b[K\n",
            "remote: Total 6743 (delta 0), reused 0 (delta 0), pack-reused 6743\u001b[K\n",
            "Receiving objects: 100% (6743/6743), 13.18 MiB | 14.72 MiB/s, done.\n",
            "Resolving deltas: 100% (4241/4241), done.\n",
            "/mmocr\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from -r requirements/build.txt (line 2)) (1.21.6)\n",
            "Collecting pyclipper\n",
            "  Downloading pyclipper-1.3.0.post3-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (604 kB)\n",
            "\u001b[K     |████████████████████████████████| 604 kB 35.1 MB/s \n",
            "\u001b[?25hRequirement already satisfied: torch>=1.1 in /usr/local/lib/python3.7/dist-packages (from -r requirements/build.txt (line 4)) (1.11.0+cu113)\n",
            "Requirement already satisfied: imgaug in /usr/local/lib/python3.7/dist-packages (from -r requirements/runtime.txt (line 1)) (0.2.9)\n",
            "Collecting lanms-neo==1.0.2\n",
            "  Downloading lanms_neo-1.0.2.tar.gz (39 kB)\n",
            "  Installing build dependencies ... \u001b[?25l\u001b[?25hdone\n",
            "  Getting requirements to build wheel ... \u001b[?25l\u001b[?25hdone\n",
            "    Preparing wheel metadata ... \u001b[?25l\u001b[?25hdone\n",
            "Requirement already satisfied: lmdb in /usr/local/lib/python3.7/dist-packages (from -r requirements/runtime.txt (line 3)) (0.99)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from -r requirements/runtime.txt (line 4)) (3.2.2)\n",
            "Collecting opencv-python!=4.5.5.*,>=4.2.0.32\n",
            "  Downloading opencv_python-4.6.0.66-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.9 MB)\n",
            "\u001b[K     |████████████████████████████████| 60.9 MB 1.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pycocotools in /usr/local/lib/python3.7/dist-packages (from -r requirements/runtime.txt (line 8)) (2.0.4)\n",
            "Collecting rapidfuzz>=2.0.0\n",
            "  Downloading rapidfuzz-2.0.15-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.8 MB)\n",
            "\u001b[K     |████████████████████████████████| 1.8 MB 56.8 MB/s \n",
            "\u001b[?25hRequirement already satisfied: scikit-image in /usr/local/lib/python3.7/dist-packages (from -r requirements/runtime.txt (line 10)) (0.18.3)\n",
            "Collecting asynctest\n",
            "  Downloading asynctest-0.13.0-py3-none-any.whl (26 kB)\n",
            "Collecting codecov\n",
            "  Downloading codecov-2.1.12-py2.py3-none-any.whl (16 kB)\n",
            "Collecting flake8\n",
            "  Downloading flake8-4.0.1-py2.py3-none-any.whl (64 kB)\n",
            "\u001b[K     |████████████████████████████████| 64 kB 3.5 MB/s \n",
            "\u001b[?25hCollecting isort\n",
            "  Downloading isort-5.10.1-py3-none-any.whl (103 kB)\n",
            "\u001b[K     |████████████████████████████████| 103 kB 70.2 MB/s \n",
            "\u001b[?25hCollecting kwarray\n",
            "  Downloading kwarray-0.6.2-py3-none-any.whl (90 kB)\n",
            "\u001b[K     |████████████████████████████████| 90 kB 13.0 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pytest in /usr/local/lib/python3.7/dist-packages (from -r requirements/tests.txt (line 7)) (3.6.4)\n",
            "Collecting pytest-cov\n",
            "  Downloading pytest_cov-3.0.0-py3-none-any.whl (20 kB)\n",
            "Collecting pytest-runner\n",
            "  Downloading pytest_runner-6.0.0-py3-none-any.whl (7.2 kB)\n",
            "Collecting ubelt\n",
            "  Downloading ubelt-1.1.1-py3-none-any.whl (177 kB)\n",
            "\u001b[K     |████████████████████████████████| 177 kB 77.8 MB/s \n",
            "\u001b[?25hCollecting xdoctest>=0.10.0\n",
            "  Downloading xdoctest-1.0.0-py3-none-any.whl (117 kB)\n",
            "\u001b[K     |████████████████████████████████| 117 kB 75.0 MB/s \n",
            "\u001b[?25hRequirement already satisfied: yapf in /usr/local/lib/python3.7/dist-packages (from -r requirements/tests.txt (line 12)) (0.32.0)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch>=1.1->-r requirements/build.txt (line 4)) (4.1.1)\n",
            "Collecting jarowinkler<1.1.0,>=1.0.3\n",
            "  Downloading jarowinkler-1.0.4-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (101 kB)\n",
            "\u001b[K     |████████████████████████████████| 101 kB 14.1 MB/s \n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from xdoctest>=0.10.0->-r requirements/tests.txt (line 11)) (1.15.0)\n",
            "Requirement already satisfied: Shapely in /usr/local/lib/python3.7/dist-packages (from imgaug->-r requirements/runtime.txt (line 1)) (1.8.2)\n",
            "Requirement already satisfied: imageio in /usr/local/lib/python3.7/dist-packages (from imgaug->-r requirements/runtime.txt (line 1)) (2.4.1)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from imgaug->-r requirements/runtime.txt (line 1)) (1.4.1)\n",
            "Requirement already satisfied: Pillow in /usr/local/lib/python3.7/dist-packages (from imgaug->-r requirements/runtime.txt (line 1)) (7.1.2)\n",
            "Requirement already satisfied: PyWavelets>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from scikit-image->-r requirements/runtime.txt (line 10)) (1.3.0)\n",
            "Requirement already satisfied: networkx>=2.0 in /usr/local/lib/python3.7/dist-packages (from scikit-image->-r requirements/runtime.txt (line 10)) (2.6.3)\n",
            "Requirement already satisfied: tifffile>=2019.7.26 in /usr/local/lib/python3.7/dist-packages (from scikit-image->-r requirements/runtime.txt (line 10)) (2021.11.2)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->-r requirements/runtime.txt (line 4)) (2.8.2)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->-r requirements/runtime.txt (line 4)) (1.4.3)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->-r requirements/runtime.txt (line 4)) (3.0.9)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->-r requirements/runtime.txt (line 4)) (0.11.0)\n",
            "Requirement already satisfied: coverage in /usr/local/lib/python3.7/dist-packages (from codecov->-r requirements/tests.txt (line 2)) (3.7.1)\n",
            "Requirement already satisfied: requests>=2.7.9 in /usr/local/lib/python3.7/dist-packages (from codecov->-r requirements/tests.txt (line 2)) (2.23.0)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.7.9->codecov->-r requirements/tests.txt (line 2)) (3.0.4)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.7.9->codecov->-r requirements/tests.txt (line 2)) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.7.9->codecov->-r requirements/tests.txt (line 2)) (2022.6.15)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.7.9->codecov->-r requirements/tests.txt (line 2)) (1.24.3)\n",
            "Collecting pyflakes<2.5.0,>=2.4.0\n",
            "  Downloading pyflakes-2.4.0-py2.py3-none-any.whl (69 kB)\n",
            "\u001b[K     |████████████████████████████████| 69 kB 9.6 MB/s \n",
            "\u001b[?25hCollecting pycodestyle<2.9.0,>=2.8.0\n",
            "  Downloading pycodestyle-2.8.0-py2.py3-none-any.whl (42 kB)\n",
            "\u001b[K     |████████████████████████████████| 42 kB 1.2 MB/s \n",
            "\u001b[?25hCollecting mccabe<0.7.0,>=0.6.0\n",
            "  Downloading mccabe-0.6.1-py2.py3-none-any.whl (8.6 kB)\n",
            "Collecting importlib-metadata<4.3\n",
            "  Downloading importlib_metadata-4.2.0-py3-none-any.whl (16 kB)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata<4.3->flake8->-r requirements/tests.txt (line 3)) (3.8.0)\n",
            "Requirement already satisfied: pluggy<0.8,>=0.5 in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (0.7.1)\n",
            "Requirement already satisfied: more-itertools>=4.0.0 in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (8.13.0)\n",
            "Requirement already satisfied: py>=1.5.0 in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (1.11.0)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (57.4.0)\n",
            "Requirement already satisfied: attrs>=17.4.0 in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (21.4.0)\n",
            "Requirement already satisfied: atomicwrites>=1.0 in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (1.4.0)\n",
            "Collecting pytest\n",
            "  Downloading pytest-7.1.2-py3-none-any.whl (297 kB)\n",
            "\u001b[K     |████████████████████████████████| 297 kB 73.6 MB/s \n",
            "\u001b[?25hCollecting coverage[toml]>=5.2.1\n",
            "  Downloading coverage-6.4.1-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl (208 kB)\n",
            "\u001b[K     |████████████████████████████████| 208 kB 74.2 MB/s \n",
            "\u001b[?25hRequirement already satisfied: tomli>=1.0.0 in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (2.0.1)\n",
            "Requirement already satisfied: iniconfig in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (1.1.1)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (21.3)\n",
            "Collecting pytest\n",
            "  Downloading pytest-7.1.1-py3-none-any.whl (297 kB)\n",
            "\u001b[K     |████████████████████████████████| 297 kB 77.6 MB/s \n",
            "\u001b[?25h  Downloading pytest-7.1.0-py3-none-any.whl (297 kB)\n",
            "\u001b[K     |████████████████████████████████| 297 kB 79.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-7.0.1-py3-none-any.whl (296 kB)\n",
            "\u001b[K     |████████████████████████████████| 296 kB 67.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-7.0.0-py3-none-any.whl (296 kB)\n",
            "\u001b[K     |████████████████████████████████| 296 kB 80.3 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.2.5-py3-none-any.whl (280 kB)\n",
            "\u001b[K     |████████████████████████████████| 280 kB 79.8 MB/s \n",
            "\u001b[?25hCollecting toml\n",
            "  Downloading toml-0.10.2-py2.py3-none-any.whl (16 kB)\n",
            "Collecting pytest\n",
            "  Downloading pytest-6.2.4-py3-none-any.whl (280 kB)\n",
            "\u001b[K     |████████████████████████████████| 280 kB 77.3 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.2.3-py3-none-any.whl (280 kB)\n",
            "\u001b[K     |████████████████████████████████| 280 kB 80.4 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.2.2-py3-none-any.whl (280 kB)\n",
            "\u001b[K     |████████████████████████████████| 280 kB 76.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.2.1-py3-none-any.whl (279 kB)\n",
            "\u001b[K     |████████████████████████████████| 279 kB 80.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.2.0-py3-none-any.whl (279 kB)\n",
            "\u001b[K     |████████████████████████████████| 279 kB 78.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.1.2-py3-none-any.whl (272 kB)\n",
            "\u001b[K     |████████████████████████████████| 272 kB 79.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.1.1-py3-none-any.whl (272 kB)\n",
            "\u001b[K     |████████████████████████████████| 272 kB 49.3 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.1.0-py3-none-any.whl (272 kB)\n",
            "\u001b[K     |████████████████████████████████| 272 kB 74.7 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.0.2-py3-none-any.whl (270 kB)\n",
            "\u001b[K     |████████████████████████████████| 270 kB 76.0 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.0.1-py3-none-any.whl (270 kB)\n",
            "\u001b[K     |████████████████████████████████| 270 kB 79.5 MB/s \n",
            "\u001b[?25h  Downloading pytest-6.0.0-py3-none-any.whl (270 kB)\n",
            "\u001b[K     |████████████████████████████████| 270 kB 75.6 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.4.3-py3-none-any.whl (248 kB)\n",
            "\u001b[K     |████████████████████████████████| 248 kB 57.0 MB/s \n",
            "\u001b[?25hRequirement already satisfied: wcwidth in /usr/local/lib/python3.7/dist-packages (from pytest->-r requirements/tests.txt (line 7)) (0.2.5)\n",
            "  Downloading pytest-5.4.2-py3-none-any.whl (247 kB)\n",
            "\u001b[K     |████████████████████████████████| 247 kB 80.0 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.4.1-py3-none-any.whl (246 kB)\n",
            "\u001b[K     |████████████████████████████████| 246 kB 80.3 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.4.0-py3-none-any.whl (247 kB)\n",
            "\u001b[K     |████████████████████████████████| 247 kB 79.3 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.3.5-py3-none-any.whl (235 kB)\n",
            "\u001b[K     |████████████████████████████████| 235 kB 79.9 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.3.4-py3-none-any.whl (235 kB)\n",
            "\u001b[K     |████████████████████████████████| 235 kB 76.6 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.3.3-py3-none-any.whl (235 kB)\n",
            "\u001b[K     |████████████████████████████████| 235 kB 80.3 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.3.2-py3-none-any.whl (234 kB)\n",
            "\u001b[K     |████████████████████████████████| 234 kB 75.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.3.1-py3-none-any.whl (233 kB)\n",
            "\u001b[K     |████████████████████████████████| 233 kB 68.9 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.3.0-py3-none-any.whl (233 kB)\n",
            "\u001b[K     |████████████████████████████████| 233 kB 76.6 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.2.4-py3-none-any.whl (227 kB)\n",
            "\u001b[K     |████████████████████████████████| 227 kB 80.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.2.3-py3-none-any.whl (227 kB)\n",
            "\u001b[K     |████████████████████████████████| 227 kB 68.1 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.2.2-py3-none-any.whl (227 kB)\n",
            "\u001b[K     |████████████████████████████████| 227 kB 75.1 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.2.1-py3-none-any.whl (226 kB)\n",
            "\u001b[K     |████████████████████████████████| 226 kB 71.6 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.2.0-py3-none-any.whl (226 kB)\n",
            "\u001b[K     |████████████████████████████████| 226 kB 78.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.1.3-py3-none-any.whl (224 kB)\n",
            "\u001b[K     |████████████████████████████████| 224 kB 75.0 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.1.2-py3-none-any.whl (224 kB)\n",
            "\u001b[K     |████████████████████████████████| 224 kB 79.6 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.1.1-py3-none-any.whl (223 kB)\n",
            "\u001b[K     |████████████████████████████████| 223 kB 77.5 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.1.0-py3-none-any.whl (223 kB)\n",
            "\u001b[K     |████████████████████████████████| 223 kB 80.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.0.1-py3-none-any.whl (221 kB)\n",
            "\u001b[K     |████████████████████████████████| 221 kB 69.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-5.0.0-py3-none-any.whl (221 kB)\n",
            "\u001b[K     |████████████████████████████████| 221 kB 69.7 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.11-py2.py3-none-any.whl (231 kB)\n",
            "\u001b[K     |████████████████████████████████| 231 kB 78.9 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.10-py2.py3-none-any.whl (231 kB)\n",
            "\u001b[K     |████████████████████████████████| 231 kB 77.5 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.9-py2.py3-none-any.whl (231 kB)\n",
            "\u001b[K     |████████████████████████████████| 231 kB 79.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.8-py2.py3-none-any.whl (230 kB)\n",
            "\u001b[K     |████████████████████████████████| 230 kB 74.9 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.7-py2.py3-none-any.whl (230 kB)\n",
            "\u001b[K     |████████████████████████████████| 230 kB 78.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.6-py2.py3-none-any.whl (230 kB)\n",
            "\u001b[K     |████████████████████████████████| 230 kB 79.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.5-py2.py3-none-any.whl (230 kB)\n",
            "\u001b[K     |████████████████████████████████| 230 kB 61.4 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.4-py2.py3-none-any.whl (229 kB)\n",
            "\u001b[K     |████████████████████████████████| 229 kB 81.2 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.3-py2.py3-none-any.whl (229 kB)\n",
            "\u001b[K     |████████████████████████████████| 229 kB 75.4 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.2-py2.py3-none-any.whl (229 kB)\n",
            "\u001b[K     |████████████████████████████████| 229 kB 72.9 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.1-py2.py3-none-any.whl (229 kB)\n",
            "\u001b[K     |████████████████████████████████| 229 kB 17.8 MB/s \n",
            "\u001b[?25h  Downloading pytest-4.6.0-py2.py3-none-any.whl (229 kB)\n",
            "\u001b[K     |████████████████████████████████| 229 kB 80.3 MB/s \n",
            "\u001b[?25hINFO: pip is looking at multiple versions of pytest-cov to determine which version is compatible with other requirements. This could take a while.\n",
            "Collecting pytest-cov\n",
            "  Downloading pytest_cov-2.12.1-py2.py3-none-any.whl (20 kB)\n",
            "  Downloading pytest_cov-2.12.0-py2.py3-none-any.whl (20 kB)\n",
            "  Downloading pytest_cov-2.11.1-py2.py3-none-any.whl (20 kB)\n",
            "  Downloading pytest_cov-2.11.0-py2.py3-none-any.whl (20 kB)\n",
            "  Downloading pytest_cov-2.10.1-py2.py3-none-any.whl (19 kB)\n",
            "  Downloading pytest_cov-2.10.0-py2.py3-none-any.whl (19 kB)\n",
            "  Downloading pytest_cov-2.9.0-py2.py3-none-any.whl (19 kB)\n",
            "Building wheels for collected packages: lanms-neo\n",
            "  Building wheel for lanms-neo (PEP 517) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for lanms-neo: filename=lanms_neo-1.0.2-cp37-cp37m-linux_x86_64.whl size=107341 sha256=7b1b939a5d496650b7e3b7f05084ac5ed4785916d0d666e9c6269680dbae7f5e\n",
            "  Stored in directory: /root/.cache/pip/wheels/09/7b/5e/1e4d24a8f94c1116afa284ce2968ef2f72b986a5457164b340\n",
            "Successfully built lanms-neo\n",
            "Installing collected packages: ubelt, pyflakes, pycodestyle, opencv-python, mccabe, jarowinkler, importlib-metadata, coverage, xdoctest, rapidfuzz, pytest-runner, pytest-cov, pyclipper, lanms-neo, kwarray, isort, flake8, codecov, asynctest\n",
            "  Attempting uninstall: opencv-python\n",
            "    Found existing installation: opencv-python 4.1.2.30\n",
            "    Uninstalling opencv-python-4.1.2.30:\n",
            "      Successfully uninstalled opencv-python-4.1.2.30\n",
            "  Attempting uninstall: importlib-metadata\n",
            "    Found existing installation: importlib-metadata 4.11.4\n",
            "    Uninstalling importlib-metadata-4.11.4:\n",
            "      Successfully uninstalled importlib-metadata-4.11.4\n",
            "  Attempting uninstall: coverage\n",
            "    Found existing installation: coverage 3.7.1\n",
            "    Uninstalling coverage-3.7.1:\n",
            "      Successfully uninstalled coverage-3.7.1\n",
            "\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
            "markdown 3.3.7 requires importlib-metadata>=4.4; python_version < \"3.10\", but you have importlib-metadata 4.2.0 which is incompatible.\n",
            "datascience 0.10.6 requires coverage==3.7.1, but you have coverage 6.4.1 which is incompatible.\n",
            "datascience 0.10.6 requires folium==0.2.1, but you have folium 0.8.3 which is incompatible.\n",
            "coveralls 0.5 requires coverage<3.999,>=3.6, but you have coverage 6.4.1 which is incompatible.\n",
            "albumentations 0.1.12 requires imgaug<0.2.7,>=0.2.5, but you have imgaug 0.2.9 which is incompatible.\u001b[0m\n",
            "Successfully installed asynctest-0.13.0 codecov-2.1.12 coverage-6.4.1 flake8-4.0.1 importlib-metadata-4.2.0 isort-5.10.1 jarowinkler-1.0.4 kwarray-0.6.2 lanms-neo-1.0.2 mccabe-0.6.1 opencv-python-4.6.0.66 pyclipper-1.3.0.post3 pycodestyle-2.8.0 pyflakes-2.4.0 pytest-cov-2.9.0 pytest-runner-6.0.0 rapidfuzz-2.0.15 ubelt-1.1.1 xdoctest-1.0.0\n",
            "Using pip 21.1.3 from /usr/local/lib/python3.7/dist-packages/pip (python 3.7)\n",
            "Value for scheme.platlib does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/lib/python3.7/dist-packages\n",
            "sysconfig: /usr/lib/python3.7/site-packages\n",
            "Value for scheme.purelib does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/lib/python3.7/dist-packages\n",
            "sysconfig: /usr/lib/python3.7/site-packages\n",
            "Value for scheme.headers does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/include/python3.7/UNKNOWN\n",
            "sysconfig: /usr/include/python3.7m/UNKNOWN\n",
            "Value for scheme.scripts does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/bin\n",
            "sysconfig: /usr/bin\n",
            "Value for scheme.data does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local\n",
            "sysconfig: /usr\n",
            "Additional context:\n",
            "user = False\n",
            "home = None\n",
            "root = None\n",
            "prefix = None\n",
            "Non-user install because site-packages writeable\n",
            "Created temporary directory: /tmp/pip-ephem-wheel-cache-5doond_x\n",
            "Created temporary directory: /tmp/pip-req-tracker-itug3_2a\n",
            "Initialized build tracking at /tmp/pip-req-tracker-itug3_2a\n",
            "Created build tracker: /tmp/pip-req-tracker-itug3_2a\n",
            "Entered build tracker: /tmp/pip-req-tracker-itug3_2a\n",
            "Created temporary directory: /tmp/pip-install-hijhv4kk\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Obtaining file:///mmocr\n",
            "  Added file:///mmocr to build tracker '/tmp/pip-req-tracker-itug3_2a'\n",
            "    Running setup.py (path:/mmocr/setup.py) egg_info for package from file:///mmocr\n",
            "    Created temporary directory: /tmp/pip-pip-egg-info-rh1_3p_y\n",
            "    Running command python setup.py egg_info\n",
            "    running egg_info\n",
            "    creating /tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info\n",
            "    writing /tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info/PKG-INFO\n",
            "    writing dependency_links to /tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info/dependency_links.txt\n",
            "    writing requirements to /tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info/requires.txt\n",
            "    writing top-level names to /tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info/top_level.txt\n",
            "    writing manifest file '/tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info/SOURCES.txt'\n",
            "    reading manifest template 'MANIFEST.in'\n",
            "    warning: no files found matching 'mmocr/.mim/model-index.yml'\n",
            "    warning: no files found matching '*.py' under directory 'mmocr/.mim/configs'\n",
            "    warning: no files found matching '*.yml' under directory 'mmocr/.mim/configs'\n",
            "    warning: no files found matching '*.sh' under directory 'mmocr/.mim/tools'\n",
            "    warning: no files found matching '*.py' under directory 'mmocr/.mim/tools'\n",
            "    adding license file 'LICENSE'\n",
            "    writing manifest file '/tmp/pip-pip-egg-info-rh1_3p_y/mmocr.egg-info/SOURCES.txt'\n",
            "  Source in /mmocr has version 0.6.0, which satisfies requirement mmocr==0.6.0 from file:///mmocr\n",
            "  Removed mmocr==0.6.0 from file:///mmocr from build tracker '/tmp/pip-req-tracker-itug3_2a'\n",
            "Requirement already satisfied: imgaug in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (0.2.9)\n",
            "Requirement already satisfied: lanms-neo==1.0.2 in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (1.0.2)\n",
            "Requirement already satisfied: lmdb in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (0.99)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (3.2.2)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (1.21.6)\n",
            "Requirement already satisfied: opencv-python!=4.5.5.*,>=4.2.0.32 in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (4.6.0.66)\n",
            "Requirement already satisfied: pyclipper in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (1.3.0.post3)\n",
            "Requirement already satisfied: pycocotools in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (2.0.4)\n",
            "Requirement already satisfied: rapidfuzz>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (2.0.15)\n",
            "Requirement already satisfied: scikit-image in /usr/local/lib/python3.7/dist-packages (from mmocr==0.6.0) (0.18.3)\n",
            "Requirement already satisfied: jarowinkler<1.1.0,>=1.0.3 in /usr/local/lib/python3.7/dist-packages (from rapidfuzz>=2.0.0->mmocr==0.6.0) (1.0.4)\n",
            "Requirement already satisfied: Pillow in /usr/local/lib/python3.7/dist-packages (from imgaug->mmocr==0.6.0) (7.1.2)\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from imgaug->mmocr==0.6.0) (1.4.1)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from imgaug->mmocr==0.6.0) (1.15.0)\n",
            "Requirement already satisfied: imageio in /usr/local/lib/python3.7/dist-packages (from imgaug->mmocr==0.6.0) (2.4.1)\n",
            "Requirement already satisfied: Shapely in /usr/local/lib/python3.7/dist-packages (from imgaug->mmocr==0.6.0) (1.8.2)\n",
            "Requirement already satisfied: PyWavelets>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from scikit-image->mmocr==0.6.0) (1.3.0)\n",
            "Requirement already satisfied: tifffile>=2019.7.26 in /usr/local/lib/python3.7/dist-packages (from scikit-image->mmocr==0.6.0) (2021.11.2)\n",
            "Requirement already satisfied: networkx>=2.0 in /usr/local/lib/python3.7/dist-packages (from scikit-image->mmocr==0.6.0) (2.6.3)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmocr==0.6.0) (1.4.3)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmocr==0.6.0) (2.8.2)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmocr==0.6.0) (0.11.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->mmocr==0.6.0) (3.0.9)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->mmocr==0.6.0) (4.1.1)\n",
            "Created temporary directory: /tmp/pip-unpack-n2sbcn5w\n",
            "Installing collected packages: mmocr\n",
            "  Value for scheme.platlib does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "  distutils: /usr/local/lib/python3.7/dist-packages\n",
            "  sysconfig: /usr/lib/python3.7/site-packages\n",
            "  Value for scheme.purelib does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "  distutils: /usr/local/lib/python3.7/dist-packages\n",
            "  sysconfig: /usr/lib/python3.7/site-packages\n",
            "  Value for scheme.headers does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "  distutils: /usr/local/include/python3.7/mmocr\n",
            "  sysconfig: /usr/include/python3.7m/mmocr\n",
            "  Value for scheme.scripts does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "  distutils: /usr/local/bin\n",
            "  sysconfig: /usr/bin\n",
            "  Value for scheme.data does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "  distutils: /usr/local\n",
            "  sysconfig: /usr\n",
            "  Additional context:\n",
            "  user = False\n",
            "  home = None\n",
            "  root = None\n",
            "  prefix = None\n",
            "  Running setup.py develop for mmocr\n",
            "    Running command /usr/bin/python3 -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '\"'\"'/mmocr/setup.py'\"'\"'; __file__='\"'\"'/mmocr/setup.py'\"'\"';f = getattr(tokenize, '\"'\"'open'\"'\"', open)(__file__) if os.path.exists(__file__) else io.StringIO('\"'\"'from setuptools import setup; setup()'\"'\"');code = f.read().replace('\"'\"'\\r\\n'\"'\"', '\"'\"'\\n'\"'\"');f.close();exec(compile(code, __file__, '\"'\"'exec'\"'\"'))' develop --no-deps\n",
            "    running develop\n",
            "    running egg_info\n",
            "    creating mmocr.egg-info\n",
            "    writing mmocr.egg-info/PKG-INFO\n",
            "    writing dependency_links to mmocr.egg-info/dependency_links.txt\n",
            "    writing requirements to mmocr.egg-info/requires.txt\n",
            "    writing top-level names to mmocr.egg-info/top_level.txt\n",
            "    writing manifest file 'mmocr.egg-info/SOURCES.txt'\n",
            "    reading manifest template 'MANIFEST.in'\n",
            "    adding license file 'LICENSE'\n",
            "    writing manifest file 'mmocr.egg-info/SOURCES.txt'\n",
            "    running build_ext\n",
            "    Creating /usr/local/lib/python3.7/dist-packages/mmocr.egg-link (link to .)\n",
            "    Adding mmocr 0.6.0 to easy-install.pth file\n",
            "\n",
            "    Installed /mmocr\n",
            "Value for scheme.platlib does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/lib/python3.7/dist-packages\n",
            "sysconfig: /usr/lib/python3.7/site-packages\n",
            "Value for scheme.purelib does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/lib/python3.7/dist-packages\n",
            "sysconfig: /usr/lib/python3.7/site-packages\n",
            "Value for scheme.headers does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/include/python3.7/UNKNOWN\n",
            "sysconfig: /usr/include/python3.7m/UNKNOWN\n",
            "Value for scheme.scripts does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local/bin\n",
            "sysconfig: /usr/bin\n",
            "Value for scheme.data does not match. Please report this to <https://github.com/pypa/pip/issues/9617>\n",
            "distutils: /usr/local\n",
            "sysconfig: /usr\n",
            "Additional context:\n",
            "user = False\n",
            "home = None\n",
            "root = None\n",
            "prefix = None\n",
            "Successfully installed mmocr-0.6.0\n",
            "Removed build tracker: '/tmp/pip-req-tracker-itug3_2a'\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: gdown in /usr/local/lib/python3.7/dist-packages (4.4.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from gdown) (3.7.1)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from gdown) (4.64.0)\n",
            "Requirement already satisfied: requests[socks] in /usr/local/lib/python3.7/dist-packages (from gdown) (2.23.0)\n",
            "Requirement already satisfied: beautifulsoup4 in /usr/local/lib/python3.7/dist-packages (from gdown) (4.6.3)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from gdown) (1.15.0)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests[socks]->gdown) (3.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests[socks]->gdown) (2022.6.15)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests[socks]->gdown) (2.10)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests[socks]->gdown) (1.24.3)\n",
            "Requirement already satisfied: PySocks!=1.5.7,>=1.5.6 in /usr/local/lib/python3.7/dist-packages (from requests[socks]->gdown) (1.7.1)\n"
          ]
        }
      ],
      "metadata": {
        "id": "DwDY3puNNmhe",
        "tags": [
          "outputPrepend"
        ],
        "outputId": "e700d8d5-7661-484c-d94d-bf7ce8cd16db",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Check Installed Dependencies Versions"
      ],
      "metadata": {
        "id": "DY64JCc0pEUu"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Check Pytorch installation\n",
        "import torch, torchvision\n",
        "print(torch.__version__, torch.cuda.is_available())\n",
        "\n",
        "# Check MMDetection installation\n",
        "import mmdet\n",
        "print(mmdet.__version__)\n",
        "\n",
        "# Check mmcv installation\n",
        "import mmcv\n",
        "from mmcv.ops import get_compiling_cuda_version, get_compiler_version\n",
        "print(mmcv.__version__)\n",
        "print(get_compiling_cuda_version())\n",
        "print(get_compiler_version())\n",
        "\n",
        "# Check mmocr installation\n",
        "import mmocr\n",
        "print(mmocr.__version__)\n",
        "\n",
        "%cd /mmocr/\n",
        "!ls"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1.11.0+cu113 True\n",
            "2.25.0\n",
            "1.5.3\n",
            "11.3\n",
            "GCC 7.3\n",
            "0.6.0\n",
            "/mmocr\n",
            "CITATION.cff  docs\t   mmocr.egg-info   requirements      setup.py\n",
            "configs       LICENSE\t   model-index.yml  requirements.txt  tests\n",
            "demo\t      MANIFEST.in  README.md\t    resources\t      tools\n",
            "docker\t      mmocr\t   README_zh-CN.md  setup.cfg\n"
          ]
        }
      ],
      "metadata": {
        "id": "JABQfPwQN52g",
        "outputId": "75381b08-9dec-4ac7-83b4-f8e132f2b933",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Perform Training on a Toy Dataset with MMOCR Recognizer\n",
        "We now demonstrate how to perform training with an MMOCR recognizer. Since training a full academic dataset is time consuming (usually takes about several hours), we will train on the toy dataset for the SAR text recognition model and visualize the predictions.\n",
        "\n",
        "Training a dataset usually consists of the following steps:\n",
        "1. Convert the dataset into a format supported by MMOCR (e.g. COCO for text detection). The annotation file can be in either .txt or .lmdb format, depending on the size of the dataset. This step is usually applicable to customized datasets, since the datasets and annotation files we provide are already in supported formats. \n",
        "2. Modify the config for training. \n",
        "3. Train the model. \n",
        "\n"
      ],
      "metadata": {
        "id": "nYon41X7RTOT"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## About the Toy Dataset\n",
        "\n",
        "In this tutorial, we will be working with the dataset under the folder `toy_dataset`. \n",
        "\n",
        "Each of the image have the accompnying `annotation` file generated using the `labelme`\n",
        "\n",
        "The images were stored under the following path\n",
        "\n",
        "`/mmocr/tests/data/toy_dataset/imgs/test`\n",
        "\n",
        "The annotation were stored under  the following path\n",
        "\n",
        "`/mmocr/tests/data/toy_dataset/labelme`\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "FElJSp1vpEUz"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Print list of images and annotation available under the`toy_dataset` folder\n",
        "!cat /mmocr/tests/data/toy_dataset/img_list.txt"
      ],
      "metadata": {
        "id": "FzyOZkrwCxEE",
        "outputId": "a988c976-fd13-483d-e4c1-7915d96d436e",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "img_10.jpg\n",
            "img_1.jpg\n",
            "img_2.jpg\n",
            "img_3.jpg\n",
            "img_4.jpg\n",
            "img_5.jpg\n",
            "img_6.jpg\n",
            "img_7.jpg\n",
            "img_8.jpg\n",
            "img_9.jpg\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "We first get a sense of what the toy dataset looks like by visualizing one of the images and labels. "
      ],
      "metadata": {
        "id": "qLEON60YC1Lj"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "import mmcv\n",
        "import matplotlib.pyplot as plt \n",
        "\n",
        "#To ensure we are in the root directory\n",
        "%cd .. \n",
        "\n",
        "img = mmcv.imread('/mmocr/tests/data/toy_dataset/imgs/test/img_1.jpg')\n",
        "\n",
        "plt.imshow(mmcv.bgr2rgb(img))\n",
        "plt.show()"
      ],
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "metadata": {
        "id": "hZfd2pnqN5-Q",
        "outputId": "212aa9c8-8bd2-481f-de68-02dc726cae32",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 258
        }
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let visualise the annotation"
      ],
      "metadata": {
        "id": "O1V3V_4L_DHQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!cat /mmocr/tests/data/toy_dataset/labelme/img_1.json"
      ],
      "metadata": {
        "id": "ZP6Y4nLbFsYv",
        "outputId": "90b59417-3a30-44f5-8d24-e8e0b23bab89",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "{\r\n",
            "  \"version\": \"5.0.1\",\r\n",
            "  \"flags\": {},\r\n",
            "  \"shapes\": [\r\n",
            "    {\r\n",
            "      \"label\": \"Genaxis Theatre\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          377,\r\n",
            "          117\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          463,\r\n",
            "          117\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          465,\r\n",
            "          130\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          378,\r\n",
            "          130\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    },\r\n",
            "    {\r\n",
            "      \"label\": \"[06]\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          493,\r\n",
            "          115\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          519,\r\n",
            "          115\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          519,\r\n",
            "          131\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          493,\r\n",
            "          131\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    },\r\n",
            "    {\r\n",
            "      \"label\": \"###\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          374,\r\n",
            "          155\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          409,\r\n",
            "          155\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          409,\r\n",
            "          170\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          374,\r\n",
            "          170\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    },\r\n",
            "    {\r\n",
            "      \"label\": \"62-03\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          492,\r\n",
            "          151\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          551,\r\n",
            "          151\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          551,\r\n",
            "          170\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          492,\r\n",
            "          170\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    },\r\n",
            "    {\r\n",
            "      \"label\": \"Carpark\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          376,\r\n",
            "          198\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          422,\r\n",
            "          198\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          422,\r\n",
            "          212\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          376,\r\n",
            "          212\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    },\r\n",
            "    {\r\n",
            "      \"label\": \"###\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          494,\r\n",
            "          190\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          539,\r\n",
            "          189\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          539,\r\n",
            "          205\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          494,\r\n",
            "          206\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    },\r\n",
            "    {\r\n",
            "      \"label\": \"###\",\r\n",
            "      \"points\": [\r\n",
            "        [\r\n",
            "          374,\r\n",
            "          1\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          494,\r\n",
            "          0\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          492,\r\n",
            "          85\r\n",
            "        ],\r\n",
            "        [\r\n",
            "          372,\r\n",
            "          86\r\n",
            "        ]\r\n",
            "      ],\r\n",
            "      \"group_id\": null,\r\n",
            "      \"shape_type\": \"polygon\",\r\n",
            "      \"flags\": {}\r\n",
            "    }\r\n",
            "  ],\r\n",
            "  \"imagePath\": \"tests/data/toy_dataset/imgs/test/img_1.jpg\",\r\n",
            "  \"imageData\": \"\",\r\n",
            "  \"imageHeight\": 720,\r\n",
            "  \"imageWidth\": 1280\r\n",
            "}"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Prepare Annotation to MMOCR format \n",
        "\n",
        "Since the annotations are in COCO format, we need to firstly convert annotation in labelme format to MMOCR's annotation format using the `labelme_converter.py`.\n",
        "\n",
        "For this particular example, we will convert the labelme format annotation in `/mmocr/tests/data/toy_dataset/labelme` to MMOCR detection labels instances_training `.txt` and export the cropped image patches for recognition task to `/mmocr/tests/data/toy_dataset/crops`. It is worth to note that the following setting would not output the `test_label.txt` as there is no test set. This information is important  when assigning  `test_anno_file1` to `train_label.txt` in subsequent step.\n",
        "\n",
        "In Google colab, it is essential to add a `!` before the Python command.\n"
      ],
      "metadata": {
        "id": "HDjdg6L8ziYS"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "For this particular dataset, preferable to use the `.jsonl` as the annotation output"
      ],
      "metadata": {
        "id": "K1gpyMVAgOg1"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!python /mmocr/tools/data/common/labelme_converter.py /mmocr/tests/data/toy_dataset/labelme /mmocr/tests/data/toy_dataset/imgs/test /mmocr/tests/data/toy_dataset --tasks recog --format jsonl"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Fur1PcDpgOCX",
        "outputId": "b1e01fd7-713f-4e92-a175-cf61dc414f57"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[                                                  ] 0/10, elapsed: 0s, ETA:\r[  ] 1/10, 78.3 task/s, elapsed: 0s, ETA:     0s\r[  ] 2/10, 86.5 task/s, elapsed: 0s, ETA:     0s\r[  ] 3/10, 90.5 task/s, elapsed: 0s, ETA:     0s\r[  ] 4/10, 93.0 task/s, elapsed: 0s, ETA:     0s\r[> ] 5/10, 94.1 task/s, elapsed: 0s, ETA:     0s\r[> ] 6/10, 93.1 task/s, elapsed: 0s, ETA:     0s\r[> ] 7/10, 91.1 task/s, elapsed: 0s, ETA:     0s\r[> ] 8/10, 91.7 task/s, elapsed: 0s, ETA:     0s\r[> ] 9/10, 91.9 task/s, elapsed: 0s, ETA:     0s\r[>>] 10/10, 91.5 task/s, elapsed: 0s, ETA:     0s\n",
            "finish\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let have a look what files has been generated from the previous calling"
      ],
      "metadata": {
        "id": "NVsw-hMkETbp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# List of crop images that being crop in accordance to the bounding box\n",
        "import os\n",
        "root = '/mmocr/tests/data/toy_dataset' # Location where the annotation and crop images are being stored\n",
        "path_crop_img=os.path.join(root,'crops')\n",
        "print([os.path.join(path, name) for path, subdirs, files in os.walk(path_crop_img) for name in files])"
      ],
      "metadata": {
        "id": "X8kvWR3UE4KG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "acfb6e1b-7f71-43e8-9bb7-4e41ec645002"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['/mmocr/tests/data/toy_dataset/crops/img_3/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_3/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_7/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_7/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_10/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_10/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_10/crop_2.jpg', '/mmocr/tests/data/toy_dataset/crops/img_6/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_6/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_2/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_2/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_9/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_9/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_1/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_1/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_1/crop_3.jpg', '/mmocr/tests/data/toy_dataset/crops/img_1/crop_2.jpg', '/mmocr/tests/data/toy_dataset/crops/img_8/crop_1.jpg', '/mmocr/tests/data/toy_dataset/crops/img_8/crop_0.jpg', '/mmocr/tests/data/toy_dataset/crops/img_8/crop_3.jpg', '/mmocr/tests/data/toy_dataset/crops/img_8/crop_2.jpg']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Lets visualise the MMOCR detection labels instances_training `.jsonl`"
      ],
      "metadata": {
        "id": "VmsBm8txF06t"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!cat /mmocr/tests/data/toy_dataset/train_label.jsonl"
      ],
      "metadata": {
        "id": "RFhKFgJZGJW-",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7322b6aa-5b58-44e5-9ca7-eb83890d010e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_7/crop_0.jpg\", \"text\": \"citi\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_7/crop_1.jpg\", \"text\": \"smrt\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_3/crop_0.jpg\", \"text\": \"fusionopolis\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_3/crop_1.jpg\", \"text\": \"fusionopolis\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_6/crop_0.jpg\", \"text\": \"Reserve\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_6/crop_1.jpg\", \"text\": \"CAUTION\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_9/crop_0.jpg\", \"text\": \"EXIT\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_9/crop_1.jpg\", \"text\": \"STAGE\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_8/crop_0.jpg\", \"text\": \"WHY\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_8/crop_1.jpg\", \"text\": \"PAY\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_8/crop_2.jpg\", \"text\": \"FOR\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_8/crop_3.jpg\", \"text\": \"NOTHING?\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_2/crop_0.jpg\", \"text\": \"EXIT\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_2/crop_1.jpg\", \"text\": \"I2R\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_1/crop_0.jpg\", \"text\": \"Genaxis Theatre\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_1/crop_1.jpg\", \"text\": \"[06]\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_1/crop_2.jpg\", \"text\": \"62-03\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_1/crop_3.jpg\", \"text\": \"Carpark\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_10/crop_0.jpg\", \"text\": \"HarbourFront\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_10/crop_1.jpg\", \"text\": \"CC22\"}\n",
            "{\"filename\": \"/mmocr/tests/data/toy_dataset/crops/img_10/crop_2.jpg\", \"text\": \"bua\"}\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!wget https://github.com/balandongiv/mmocr_tutorial/blob/main/my_cfg.py.txt"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0IgGFz00c52e",
        "outputId": "eb235a29-0fba-458f-ad56-324c05529d2f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2022-06-27 04:06:29--  https://github.com/balandongiv/mmocr_tutorial/blob/main/my_cfg.py.txt\n",
            "Resolving github.com (github.com)... 140.82.112.4\n",
            "Connecting to github.com (github.com)|140.82.112.4|:443... connected.\n",
            "HTTP request sent, awaiting response... 404 Not Found\n",
            "2022-06-27 04:06:30 ERROR 404: Not Found.\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Configuration file\n",
        "\n",
        "The MMOCR's config system incorporate modular and inheritance design, which is convenient to conduct various experiments. \n",
        "\n",
        "**REMARK**\n",
        "The modular code in each cell is for explaination purpose, toward the end, the code will be paste in a `.py` file.\n",
        "\n",
        "\n",
        "The config file structure consist on the parameter \n",
        "1. [model](https://mmocr.readthedocs.io/en/latest/tutorials/config.html#model),\n",
        "2. [data & pipeline](https://mmocr.readthedocs.io/en/latest/tutorials/config.html#data-pipeline), \n",
        "3. [training schedule](https://mmocr.readthedocs.io/en/latest/tutorials/config.html#training-schedule), and \n",
        "4. [runtime setting](https://mmocr.readthedocs.io/en/latest/tutorials/config.html#runtime-setting). \n",
        "\n",
        "Please refer to [config file structure](https://mmocr.readthedocs.io/en/latest/tutorials/config.html#config-file-structure) for detailed documentation"
      ],
      "metadata": {
        "id": "i-GrV0xSkAc3"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Firstly, We import the four predefine configurations of the `model`, `data $ pipeline`, `training schedule` and `runtime setting` as follow."
      ],
      "metadata": {
        "id": "bazeIZ60zvjA"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "```\n",
        "_base_ = [\n",
        "    '/mmocr/configs/_base_/recog_models/sar.py',\n",
        "    '/mmocr/configs/_base_/recog_pipelines/sar_pipeline.py',\n",
        "    '/mmocr/configs/_base_/schedules/schedule_adam_step_5e.py',\n",
        "    '/mmocr/configs/_base_/default_runtime.py',\n",
        "]\n",
        "```\n",
        "\n",
        "Notice that all files were configured under the `.py` extension. It is important to note that the `.py` cannot contain `modular` such as `import os`.\n",
        "\n",
        "\n",
        "The pipeline can be assigned as \n",
        "\n",
        "\n",
        "\n",
        "```\n",
        "train_pipeline = {{_base_.train_pipeline}}\n",
        "test_pipeline = {{_base_.test_pipeline}}\n",
        "```\n",
        "\n"
      ],
      "metadata": {
        "id": "FCjW6K2932zk"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Define the training and validation datasets.\n",
        "\n",
        "In this tutorial, we are using the `crop` images ( `mmocr/tests/data/toy_dataset/crops`) and its annotation (`mmocr/tests/data/toy_dataset/train_label.jsonl`) as a training dataset.\n",
        "\n",
        "The training dataset can be configured as;\n",
        "\n",
        "```\n",
        "dataset_type = 'OCRDataset'\n",
        "root = '/mmocr/tests/data/toy_dataset' # Location where the annotation and crop images are being stored\n",
        "\n",
        "img_prefix ='/mmocr/tests/data/toy_dataset/crops'\n",
        "train_anno_file1 = '/mmocr/tests/data/toy_dataset/train_label.jsonl'\n",
        "\n",
        "\n",
        "loader_dt_train = dict(type='AnnFileLoader',\n",
        "                            repeat=100,                   \n",
        "                            file_format='txt',  # only txt and lmdb\n",
        "                            file_storage_backend='disk',\n",
        "                            parser=dict(type='LineJsonParser',\n",
        "                                        keys=['filename', 'text']))\n",
        "\n",
        "train_datasets1 = dict(type='OCRDataset',\n",
        "                       img_prefix=img_prefix,\n",
        "                       ann_file=train_anno_file1,\n",
        "                       loader=loader_dt_train,\n",
        "                       pipeline=None,           \n",
        "                       test_mode=False)\n",
        "\n",
        "\n",
        "train_list = [train_datasets1]\n"
      ],
      "metadata": {
        "id": "azZObd1Z0raJ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "We also need to configure the `testing` dataset. For simplicity, we assume the training and test dataset are the same \n",
        "\n",
        "```\n",
        "loader_dt_val = dict(type='AnnFileLoader',\n",
        "                            repeat=1,                   \n",
        "                            file_format='txt',  # only txt and lmdb\n",
        "                            file_storage_backend='disk',\n",
        "                            parser=dict(type='LineJsonParser',\n",
        "                                        keys=['filename', 'text']))\n",
        "\n",
        "train_datasets1 = dict(type='OCRDataset',\n",
        "                       img_prefix=img_prefix,\n",
        "                       ann_file=train_anno_file1,\n",
        "                       loader=loader_dt_val,\n",
        "                       pipeline=None,           \n",
        "                       test_mode=False)\n",
        "\n",
        "\n",
        "test_list =  [train_datasets1]\n",
        "```"
      ],
      "metadata": {
        "id": "SPEXbwvLpvKs"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "More detail about [repeat](https://github.com/open-mmlab/mmocr/issues/858#issuecomment-1166156290)"
      ],
      "metadata": {
        "id": "fl1Xc73x0lMB"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "The pipeline for `train` and `test` can be configured as below\n",
        "\n",
        "\n",
        "\n",
        "```\n",
        "train_pipeline = {{_base_.train_pipeline}}\n",
        "test_pipeline = {{_base_.test_pipeline}}\n",
        "```\n",
        "\n"
      ],
      "metadata": {
        "id": "5EwK10-xqez7"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "We also need to specify to direectory to save all files and logs.\n",
        "\n",
        "`work_dir = '/mmocr/demo/tutorial_exps'`"
      ],
      "metadata": {
        "id": "P02c3JSLpYHJ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Finally, assign all the informations into the `data` dictionary. It is worth to note, this setting is almost similar to the build-in [script](https://github.com/open-mmlab/mmocr/blob/72a79f9350277566caae2fdf3ab9cd614f313d28/configs/textrecog/sar/sar_r31_parallel_decoder_toy_dataset.py#L14).\n",
        "\n",
        "\n",
        "\n",
        "```\n",
        "data = dict(\n",
        "    workers_per_gpu=2,\n",
        "    samples_per_gpu=8,\n",
        "    train=dict(\n",
        "        type='UniformConcatDataset',\n",
        "        datasets=train_list,\n",
        "        pipeline=train_pipeline),\n",
        "    val=dict(\n",
        "        type='UniformConcatDataset',\n",
        "        datasets=test_list,\n",
        "        pipeline=test_pipeline),\n",
        "    test=dict(\n",
        "        type='UniformConcatDataset',\n",
        "        datasets=test_list,\n",
        "        pipeline=test_pipeline)\n",
        "    )\n",
        "\n",
        "evaluation = dict(interval=1, metric='acc')\n",
        "```"
      ],
      "metadata": {
        "id": "00GJwwJQ5h3H"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "As mentioned, this `cfg` need to be stored in a `.py` file.\n",
        "For simplicity, lets download the `.py` file and stored at `/content/my_cfg.py`."
      ],
      "metadata": {
        "id": "8WKFkf1Z2e6l"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import gdown\n",
        "gdown.download(url='https://drive.google.com/file/d/1mv3JSGfPhzlSzCRPN9Ddsd_9mQILLhVV/view?usp=sharing', output='/content/', quiet=True, fuzzy=True)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "id": "r-ELyw0Bg6Qb",
        "outputId": "653e3f6c-64cf-4e25-9b37-f8de72335c45"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'/content/my_cfg.py'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Lets create the configration file"
      ],
      "metadata": {
        "id": "Tp2_REBignfH"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from mmcv import Config\n",
        "\n",
        "cfg = Config.fromfile('/content/my_cfg.py')"
      ],
      "metadata": {
        "id": "CmQgFAV_9783"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "It is also possible to modify the `cfg` post creation.\n",
        "\n",
        "In order to perform inference for carnn on colab, we need to adjust the config values to accommodate some of the settings of colab such as the number of GPU available. "
      ],
      "metadata": {
        "id": "MLDAKp5cA-st"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from mmdet.apis import set_random_seed\n",
        "\n",
        "\n",
        "# The original learning rate (LR) is set for 8-GPU training.\n",
        "# We divide it by 8 since we only use one GPU.\n",
        "# For detail discussion See https://github.com/open-mmlab/mmocr/issues/681#issuecomment-1009526228 \n",
        "cfg.optimizer = dict(type = 'Adam', \n",
        "                 lr = 0.001 / 8)\n",
        "\n",
        "\n",
        "cfg.lr_config = dict(policy = 'step', \n",
        "                 step = [3, 4], \n",
        "                 warmup = None  # Set warmup to None when work in Colab\n",
        "                 ) \n",
        "\n",
        "\n",
        "cfg.log_config = dict(\n",
        "                  interval = 5,  # Choose to log training results every 40 images to reduce the size of log file. \n",
        "                  hooks = [dict(type = 'TextLoggerHook')])\n",
        "\n",
        "seed = 0\n",
        "set_random_seed(0, deterministic=False)\n",
        "cfg.gpu_ids = range(1)\n",
        "\n",
        "# Let also limit the training to only 6 epochs\n",
        "\n",
        "cfg.runner = dict(type='EpochBasedRunner', max_epochs=6) \n",
        "\n",
        "#save the model every 2 epoch\n",
        "cfg.checkpoint_config = dict(interval=2) \n",
        "# print(f'Config:\\n{cfg.pretty_text}')"
      ],
      "metadata": {
        "id": "eY1-3B7PBHRV"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let also change the [dict_type](https://github.com/open-mmlab/mmocr/issues/1113) to recognize characters from `0~9` , `A~Z` and `a-z` only and `no space`,"
      ],
      "metadata": {
        "id": "rG246XrEhZrJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Please Ignore this, This is for troubleshooting purpose only\n",
        "# DICT62=['0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ'] # 62 characters\n",
        "\n",
        "# \"\"\"\n",
        "# 'BaseConvertor', 'CTCConvertor', 'AttnConvertor', 'SegConvertor','ABIConvertor'\n",
        "# \"\"\"\n",
        "\n",
        "# # Original value\n",
        "# # label_convertor = dict(type='AttnConvertor', dict_type='DICT90', with_unknown=True)\n",
        "\n",
        "# cfg.label_convertor = dict(\n",
        "#     type='AttnConvertor', dict_type=None,dict_list=DICT62, with_unknown=False,lower=False)\n",
        "\n",
        "# cfg.model.label_convertor=cfg.label_convertor\n",
        "\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "t0xgiJSphbx7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Train the SAR Text Recognizer \n",
        "Finally, we train the SAR text recognizer on the toy dataset for five epochs. \n",
        "\n",
        "This steps took about ~ 5 minutes to complete or <15 if Colab is connected with GPU \n"
      ],
      "metadata": {
        "id": "TZj5vyqEmulE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from mmocr.datasets import build_dataset\n",
        "from mmocr.models import build_detector\n",
        "from mmocr.apis import train_detector\n",
        "import os.path as osp\n",
        "\n",
        "\n",
        "# Build dataset\n",
        "datasets = [build_dataset(cfg.data.train)]\n",
        "\n",
        "# # Build the detector\n",
        "model = build_detector(cfg.model, train_cfg=cfg.get('train_cfg'), test_cfg=cfg.get('test_cfg'))\n",
        "\n",
        "# # Add an attribute for visualization convenience\n",
        "model.CLASSES = datasets[0].CLASSES\n",
        "\n",
        "# # Create work_dir\n",
        "mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))\n",
        "train_detector(model, datasets, cfg, distributed=False, validate=True)"
      ],
      "metadata": {
        "id": "aGtuC29ys8Ej",
        "cellView": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "7a6c0e58-41e1-49a4-8bf1-4ca31874605a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/mmocr/mmocr/apis/utils.py:53: UserWarning: Remove \"MultiRotateAugOCR\" to support batch inference since samples_per_gpu > 1.\n",
            "  warnings.warn(warning_msg)\n",
            "2022-06-27 04:07:51,299 - mmocr - INFO - Start running, host: root@0911d87d2aae, work_dir: /mmocr/demo/tutorial_exps\n",
            "2022-06-27 04:07:51,305 - mmocr - INFO - Hooks will be executed in the following order:\n",
            "before_run:\n",
            "(VERY_HIGH   ) StepLrUpdaterHook                  \n",
            "(NORMAL      ) CheckpointHook                     \n",
            "(NORMAL      ) EvalHook                           \n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "before_train_epoch:\n",
            "(VERY_HIGH   ) StepLrUpdaterHook                  \n",
            "(NORMAL      ) EvalHook                           \n",
            "(LOW         ) IterTimerHook                      \n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "before_train_iter:\n",
            "(VERY_HIGH   ) StepLrUpdaterHook                  \n",
            "(NORMAL      ) EvalHook                           \n",
            "(LOW         ) IterTimerHook                      \n",
            " -------------------- \n",
            "after_train_iter:\n",
            "(ABOVE_NORMAL) OptimizerHook                      \n",
            "(NORMAL      ) CheckpointHook                     \n",
            "(NORMAL      ) EvalHook                           \n",
            "(LOW         ) IterTimerHook                      \n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "after_train_epoch:\n",
            "(NORMAL      ) CheckpointHook                     \n",
            "(NORMAL      ) EvalHook                           \n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "before_val_epoch:\n",
            "(LOW         ) IterTimerHook                      \n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "before_val_iter:\n",
            "(LOW         ) IterTimerHook                      \n",
            " -------------------- \n",
            "after_val_iter:\n",
            "(LOW         ) IterTimerHook                      \n",
            " -------------------- \n",
            "after_val_epoch:\n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "after_run:\n",
            "(VERY_LOW    ) TextLoggerHook                     \n",
            " -------------------- \n",
            "2022-06-27 04:07:51,307 - mmocr - INFO - workflow: [('train', 1)], max: 6 epochs\n",
            "2022-06-27 04:07:51,313 - mmocr - INFO - Checkpoints will be saved to /mmocr/demo/tutorial_exps by HardDiskBackend.\n",
            "/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:780: UserWarning: Note that order of the arguments: ceil_mode and return_indices will changeto match the args list in nn.MaxPool2d in a future release.\n",
            "  warnings.warn(\"Note that order of the arguments: ceil_mode and return_indices will change\"\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "Exception",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mException\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-16-4db527398459>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     16\u001b[0m \u001b[0;31m# # Create work_dir\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     17\u001b[0m \u001b[0mmmcv\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmkdir_or_exist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mosp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mabspath\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwork_dir\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 18\u001b[0;31m \u001b[0mtrain_detector\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdatasets\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcfg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdistributed\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalidate\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m/mmocr/mmocr/apis/train.py\u001b[0m in \u001b[0;36mtrain_detector\u001b[0;34m(model, dataset, cfg, distributed, validate, timestamp, meta)\u001b[0m\n\u001b[1;32m    153\u001b[0m     \u001b[0;32melif\u001b[0m \u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mload_from\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    154\u001b[0m         \u001b[0mrunner\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mload_checkpoint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mload_from\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 155\u001b[0;31m     \u001b[0mrunner\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_loaders\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcfg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mworkflow\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    156\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    157\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/mmcv/runner/epoch_based_runner.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, data_loaders, workflow, max_epochs, **kwargs)\u001b[0m\n\u001b[1;32m    128\u001b[0m                     \u001b[0;32mif\u001b[0m \u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'train'\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mepoch\u001b[0m \u001b[0;34m>=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_max_epochs\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    129\u001b[0m                         \u001b[0;32mbreak\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 130\u001b[0;31m                     \u001b[0mepoch_runner\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_loaders\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    131\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    132\u001b[0m         \u001b[0mtime\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msleep\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# wait for some hooks like loggers to finish\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/mmcv/runner/epoch_based_runner.py\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(self, data_loader, **kwargs)\u001b[0m\n\u001b[1;32m     49\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_inner_iter\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mi\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     50\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall_hook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'before_train_iter'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 51\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_iter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_batch\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_mode\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     52\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall_hook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'after_train_iter'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     53\u001b[0m             \u001b[0;32mdel\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata_batch\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/mmcv/runner/epoch_based_runner.py\u001b[0m in \u001b[0;36mrun_iter\u001b[0;34m(self, data_batch, train_mode, **kwargs)\u001b[0m\n\u001b[1;32m     28\u001b[0m         \u001b[0;32melif\u001b[0m \u001b[0mtrain_mode\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     29\u001b[0m             outputs = self.model.train_step(data_batch, self.optimizer,\n\u001b[0;32m---> 30\u001b[0;31m                                             **kwargs)\n\u001b[0m\u001b[1;32m     31\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     32\u001b[0m             \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mval_step\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_batch\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moptimizer\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/mmcv/parallel/data_parallel.py\u001b[0m in \u001b[0;36mtrain_step\u001b[0;34m(self, *inputs, **kwargs)\u001b[0m\n\u001b[1;32m     73\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     74\u001b[0m         \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mscatter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice_ids\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 75\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodule\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_step\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     76\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     77\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mval_step\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/mmocr/mmocr/models/textrecog/recognizer/base.py\u001b[0m in \u001b[0;36mtrain_step\u001b[0;34m(self, data, optimizer)\u001b[0m\n\u001b[1;32m    156\u001b[0m                 DDP model, num_samples refers to the batch size for each GPU).\n\u001b[1;32m    157\u001b[0m         \"\"\"\n\u001b[0;32m--> 158\u001b[0;31m         \u001b[0mlosses\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    159\u001b[0m         \u001b[0mloss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlog_vars\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_parse_losses\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlosses\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    160\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m   1108\u001b[0m         if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks\n\u001b[1;32m   1109\u001b[0m                 or _global_forward_hooks or _global_forward_pre_hooks):\n\u001b[0;32m-> 1110\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mforward_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1111\u001b[0m         \u001b[0;31m# Do not call functions when jit is used\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1112\u001b[0m         \u001b[0mfull_backward_hooks\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnon_full_backward_hooks\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/mmcv/runner/fp16_utils.py\u001b[0m in \u001b[0;36mnew_func\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m    114\u001b[0m                                 f'method of those classes {supported_types}')\n\u001b[1;32m    115\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mhasattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'fp16_enabled'\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfp16_enabled\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 116\u001b[0;31m                 \u001b[0;32mreturn\u001b[0m \u001b[0mold_func\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    117\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    118\u001b[0m             \u001b[0;31m# get the arg spec of the decorated method\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/mmocr/mmocr/models/textrecog/recognizer/base.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, img, img_metas, return_loss, **kwargs)\u001b[0m\n\u001b[1;32m     82\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     83\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mreturn_loss\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 84\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward_train\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mimg_metas\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     85\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     86\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/mmocr/mmocr/models/textrecog/recognizer/encode_decode_recognizer.py\u001b[0m in \u001b[0;36mforward_train\u001b[0;34m(self, img, img_metas)\u001b[0m\n\u001b[1;32m    101\u001b[0m         \u001b[0mgt_labels\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mimg_meta\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'text'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mimg_meta\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mimg_metas\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    102\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 103\u001b[0;31m         \u001b[0mtargets_dict\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlabel_convertor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstr2tensor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgt_labels\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    104\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    105\u001b[0m         \u001b[0mout_enc\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/mmocr/mmocr/models/textrecog/convertors/attn.py\u001b[0m in \u001b[0;36mstr2tensor\u001b[0;34m(self, strings)\u001b[0m\n\u001b[1;32m     87\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     88\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpadded_targets\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 89\u001b[0;31m         \u001b[0mindexes\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstr2idx\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstrings\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     90\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mindex\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mindexes\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     91\u001b[0m             \u001b[0mtensor\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mLongTensor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindex\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/mmocr/mmocr/models/textrecog/convertors/base.py\u001b[0m in \u001b[0;36mstr2idx\u001b[0;34m(self, strings)\u001b[0m\n\u001b[1;32m     80\u001b[0m                 \u001b[0mchar_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mchar2idx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mchar\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munknown_idx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     81\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mchar_idx\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 82\u001b[0;31m                     raise Exception(f'Chararcter: {char} not in dict,'\n\u001b[0m\u001b[1;32m     83\u001b[0m                                     \u001b[0;34mf' please check gt_label and use'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     84\u001b[0m                                     \u001b[0;34mf' custom dict file,'\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mException\u001b[0m: Chararcter: S not in dict, please check gt_label and use custom dict file, or set \"with_unknown=True\""
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Test and Visualize the Predictions\n",
        "\n",
        "For completeness, we also perform testing on the latest checkpoint and evaluate the results with hmean-iou metrics. The predictions are saved in the ./outputs file. "
      ],
      "metadata": {
        "id": "sklydRNXnfJk"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Visualize the testing image.\n",
        "img='/mmocr/tests/data/toy_dataset/imgs/test/img_1.jpg' # This should be an overfit, since we testing an image that has been used for training!\n",
        "\n",
        "_img = mmcv.imread(img)\n",
        "plt.figure(figsize=(4, 4))\n",
        "plt.imshow(mmcv.bgr2rgb(_img))\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "UcaZm7qY3yGT"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(cfg.work_dir)"
      ],
      "metadata": {
        "id": "zFRom8nBEppE"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "from mmocr.apis import init_detector, model_inference\n",
        "\n",
        "\n",
        "img='/mmocr/tests/data/toy_dataset/imgs/test/img_1.jpg'  # This should be an overfit, since we testing an image that has been used for training!\n",
        "\n",
        "\n",
        "\n",
        "checkpoint=os.path.join(cfg.work_dir,'epoch_5.pth')\n",
        "\n",
        "\n",
        "out_file = 'outputs/1036169s.jpg'   # Rendered image save location\n",
        "\n",
        "model = init_detector(cfg, checkpoint, device=\"cuda:0\")\n",
        "if model.cfg.data.test['type'] == 'ConcatDataset':\n",
        "    model.cfg.data.test.pipeline = model.cfg.data.test['datasets'][0].pipeline\n",
        "\n",
        "# Lets do inference to the image using the model that we just train.\n",
        "result = model_inference(model, img)\n",
        "print(f'result: {result}')\n",
        "\n",
        "img = model.show_result(img, result, \n",
        "                        out_file=out_file, \n",
        "                        show=False)\n",
        "\n",
        "mmcv.imwrite(img, out_file)"
      ],
      "outputs": [],
      "metadata": {
        "id": "-HbXY7uUpEU1"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "source": [
        "# Visualize the results\n",
        "predicted_img = mmcv.imread('./outputs/1036169s.jpg')\n",
        "plt.figure(figsize=(4, 4))\n",
        "plt.imshow(mmcv.bgr2rgb(predicted_img))\n",
        "plt.show()"
      ],
      "outputs": [],
      "metadata": {
        "id": "k3s27QIGQCnT"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Resume Training from existing training point\n"
      ],
      "metadata": {
        "id": "xzVev3FMGc3q"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from mmcv import Config\n",
        "from mmdet.apis import set_random_seed\n",
        "cfg = Config.fromfile('/content/my_cfg.py')\n",
        "\n",
        "\n",
        "cfg.gpu_ids = range(1)\n",
        "\n",
        "# Let also limit the training to only 12 epochs\n",
        "\n",
        "cfg.runner = dict(type='EpochBasedRunner', max_epochs=20) \n",
        "\n",
        "#save the model every 2 epoch\n",
        "cfg.checkpoint_config = dict(interval=5) # Default the interval was set to 100\n",
        "\n",
        "print(f'Config:\\n{cfg.pretty_text}')"
      ],
      "metadata": {
        "id": "9beDV1xAR-3H"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "mmocr offer the capability to `resume` training.\n",
        "\n",
        "\n",
        "\n",
        "or use `pre-trained` model which will be used to initialize the network paramters.\n",
        "\n",
        "Please select or set either `resume_from` or `load_from`"
      ],
      "metadata": {
        "id": "7qtT_MNrCGfz"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "To ` resume` training"
      ],
      "metadata": {
        "id": "2SA6abX9CvXQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Resume training from a previously saved checkpoint, which will inherit the training epoch and optimizer parameters.\n",
        "cfg.resume_from='/mmocr/demo/tutorial_exps/epoch_6.pth'"
      ],
      "metadata": {
        "id": "37ZPRjmhCyAG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "To use a `pre-trained` model"
      ],
      "metadata": {
        "id": "yzXo0fxMDAgF"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Path to the pre-trained model\n",
        "cfg.load_from='/mmocr/demo/tutorial_exps/epoch_6.pth'"
      ],
      "metadata": {
        "id": "EXZ3Zzh_DFop"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Interesting discussion from MMOCR github"
      ],
      "metadata": {
        "id": "hTwUGnRP24I3"
      }
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "6zzpkX4k4y1H"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "1.   Turn on imshow for [debugging](https://github.com/open-mmlab/mmocr/issues/410#issuecomment-892480963)\n",
        "2.   Different methods support differnet number of classes [(tokens)](https://github.com/open-mmlab/mmocr/issues/424#issuecomment-896568710). For example, SAR is DICT90 while CRNN is DICT36. But for evaluation metric, they are the same.\n",
        "3. `word_acc_ignore_case_symbol` is the [results that removing some tokens](https://github.com/open-mmlab/mmocr/issues/424#issuecomment-896574323) that are not a number or English characters.\n",
        "4. Evaluating on DICT36 is common sense in the academic community because it can get a [higher accuracy](https://github.com/open-mmlab/mmocr/issues/424#issuecomment-1086852698).\n",
        "5. Problems in training MJ + ST training set using robustscanner algorithm ( [Issue](https://github.com/open-mmlab/mmocr/issues/508))\n",
        "6. [visualize loss curve](https://github.com/open-mmlab/mmocr/issues/514).\n",
        "7. [Finetuning a model generally](https://github.com/open-mmlab/mmocr/issues/565#issuecomment-963916718) requires a much smaller learning rate than training from scratch. And it's hard to finetune the model for special cases, especially when your data is not large enough and your model is just likely to overfit.\n",
        "8. To extract only metric results, [batch testing would be enough](https://github.com/open-mmlab/mmocr/issues/594#issuecomment-971176307).\n",
        "9. char_precision is the average percentage of predicted characters that correctly match the ground truth texts. 1-N.E.D is an official metric used in the ICDAR competition where N.E.D stands for Normalized Edit Distance. The motivation of adaptation is that \"Chinese script has much more and also usually longer vocabulary than Latin script, which makes word accuracy metric too harsh to proper evaluate T2.2\". (read https://rrc.cvc.uab.es/?ch=14&com=tasks). And it's better to read through their implementation [REF](https://github.com/open-mmlab/mmocr/issues/760#issuecomment-1034664840)\n",
        "\n",
        "10. How to handle space. [Trick 1](https://github.com/open-mmlab/mmocr/issues/766#issuecomment-1030861493), or How [Create OCR for only number?](https://github.com/open-mmlab/mmocr/issues/766)\n",
        "\n",
        "`label_convertor = dict(\n",
        "    type='CTCConvertor', dict_list=list('0123456789 '), with_unknown=False)`\n",
        "\n",
        "\n",
        "11. label.txt and label.lmdb contain exactly the same data in two different formats. This toy dataset demo is designed to demonstrate the usage of different data formats, and how to concat multiple datasets. For real applications, you only need to prepare the data annotations in one format (.txt or .lmdb).[ref](https://github.com/open-mmlab/mmocr/issues/817#issuecomment-1061543477)\n",
        "\n",
        "11. Diffrence difference between the 'load_from' and 'resume_from' . A checkpoint file contains not only the weights of the model but metadata of the training state, e.g., the current iteration # and the learning rate. --load-from only loads the weights from the checkpoint and starts training from the very first iteration, whereas --resume-from loads additional metadata to resume exactly from the stage where the checkpoint was saved [ref](https://github.com/open-mmlab/mmocr/issues/839#issuecomment-1067462651).\n",
        "12. Use case to repeat the training on a train dataset for x repeat of times?>> Sometimes this feature is needed when we train a model on a set of datasets with imbalanced sizes, where tiling the small dataset several times is the most straightforward way to alleviate the bias brought by the large ones. [SAR](https://mmocr.readthedocs.io/en/latest/textrecog_models.html#sar) is an example [ref](https://github.com/open-mmlab/mmocr/issues/858#issuecomment-1166793020)."
      ],
      "metadata": {
        "id": "G_4pGd_F3DsX"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Discussion about learning policy\n",
        "\n",
        "0. Learning policy [example](https://github.com/open-mmlab/mmocr/blob/main/configs/_base_/schedules/schedule_adam_600e.py) \n",
        "1. Passing a list of steps to step policy means that lr only decays at your specific steps. You should pass an integer X instead so that lr would decay every X steps. [Ref](https://github.com/open-mmlab/mmocr/issues/670#issuecomment-999991224)\n",
        "2. What is the learning policy on text detection models and what is set and with what values?   Learning policy only determines how the learning rate [changes over time](https://github.com/open-mmlab/mmocr/issues/681#issuecomment-1001285752),  passing 'poly' to `lr_config' will call PolyLrUpdaterHook to apply its policy.\n",
        "3. [Change lp according to GPU number](https://github.com/open-mmlab/mmocr/issues/681#issuecomment-1009526228). . According to the [metafile](https://github.com/open-mmlab/mmocr/blob/dab5725d70b92ada4383565c3e60f6634616fd60/configs/textdet/textsnake/metafile.yml), TextSnake was trained on 8 GPUs with the actual batch size 32 (samples_per_gpus * num_gpus). Even if you want to scale down the learning rate in proportion to how your actual batch size change, the learning rate should not be less than 0.007/(32 / 8)=0.00175.\n"
      ],
      "metadata": {
        "id": "wIE-TARf6bl4"
      }
    }
  ]
}