Add Docker Env and web demo through Cog

README.md

@@ -3,6 +3,15 @@ This is a reference implementation of 3D Ken Burns Effect from a Single Image [1
 
 <a href="https://arxiv.org/abs/1909.05483" rel="Paper"><img src="http://content.sniklaus.com/kenburns/paper.jpg" alt="Paper" width="100%"></a>
 
+
+## demo 
+Click the link below to run inference through Replicate's web demo:
+
+[Demo and Docker image on Replicate](https://replicate.com/sniklaus/3d-ken-burns)
+
+
+<a href="https://replicate.com/sniklaus/3d-ken-burns"><img src="https://replicate.com/sniklaus/3d-ken-burns/badge"></a>
+
 ## setup
 Several functions are implemented in CUDA using CuPy, which is why CuPy is a required dependency. It can be installed using `pip install cupy` or alternatively using one of the provided [binary packages](https://docs.cupy.dev/en/stable/install.html#installing-cupy) as outlined in the CuPy repository. Please also make sure to have the `CUDA_HOME` environment variable configured.
 
@@ -29,6 +38,7 @@ python depthestim.py --in ./images/doublestrike.jpg --out ./depthestim.npy
 
 To benchmark the depth estimation, run `python benchmark-ibims.py` or `python benchmark-nyu.py`. You can use it to easily verify that the provided implementation runs as expected.
 
+
 ## colab
 If you do not have a suitable environment to run this projects then you could give Colab a try. It allows you to run the project in the cloud, free of charge. There are several people who provide Colab notebooks that should get you started. A few that I am aware of include one from [Arnaldo Gabriel](https://colab.research.google.com/github/agmm/colab-3d-ken-burns/blob/master/automatic-3d-ken-burns.ipynb), one from [Vlad Alex](https://towardsdatascience.com/very-spatial-507aa847179d), and one from [Ahmed Harmouche](https://github.com/wpmed92/3d-ken-burns-colab).
 
@@ -106,4 +116,4 @@ This is a project by Adobe Research. It is licensed under the [Creative Commons
 ```
 
 ## acknowledgment
-The video above uses materials under a Creative Common license or with the owner's permission, as detailed at the end.
+The video above uses materials under a Creative Common license or with the owner's permission, as detailed at the end.

cog.yaml

@@ -0,0 +1,31 @@
+# Configuration for Cog ⚙️
+# Reference: https://github.com/replicate/cog/blob/main/docs/yaml.md
+
+build:
+  # set to true if your model requires a GPU
+  gpu: true
+  cuda: "11.4"
+
+  # python version in the form '3.8' or '3.8.12'
+  python_version: "3.8"
+
+  # a list of packages in the format <package-name===<version>
+  python_packages:
+
+    - "ipython==7.33.0"
+    - "flask==2.0.3"
+    - "gevent==1.3.0"
+    - "moviepy==1.0.0"
+    - "numpy==1.21.6"
+    - "h5py==3.7.0"
+    - "torch==1.11.0"
+    - "torchvision==0.12.0"
+
+  # commands run after the environment is setup
+  run:
+    - "apt-get update && apt-get install -y python3-opencv"
+    - "pip install opencv-python"
+    - "pip install scipy"
+    - "pip install cupy-cuda114 --pre"
+# predict.py defines how predictions are run on your model
+predict: "predict.py:Predictor"

predict.py

@@ -0,0 +1,130 @@
+""" Cog-based inference for animating still image (3D Ken Burns effect)"""
+# Prediction interface for Cog ⚙️
+# https://github.com/replicate/cog/blob/main/docs/python.md
+
+import tempfile
+import warnings
+
+from cog import BasePredictor, Input, Path
+
+warnings.filterwarnings("ignore")
+
+#!/usr/bin/env python
+
+import base64
+import getopt
+import glob
+import io
+import math
+import os
+import random
+import re
+import shutil
+import sys
+import tempfile
+import time
+import urllib
+import zipfile
+
+import cupy
+import cv2
+import flask
+import gevent
+import gevent.pywsgi
+import h5py
+import moviepy
+import moviepy.editor
+import numpy
+import scipy
+import scipy.io
+import torch
+import torchvision
+
+##########################################################
+
+assert (
+    int(str("").join(torch.__version__.split(".")[0:2])) >= 12
+)  # requires at least pytorch version 1.2.0
+
+torch.set_grad_enabled(
+    False
+)  # make sure to not compute gradients for computational performance
+
+torch.backends.cudnn.enabled = (
+    True  # make sure to use cudnn for computational performance
+)
+
+##########################################################
+
+objCommon = {}
+
+exec(open("./common.py", "r").read())
+
+exec(open("./models/disparity-estimation.py", "r").read())
+exec(open("./models/disparity-adjustment.py", "r").read())
+exec(open("./models/disparity-refinement.py", "r").read())
+exec(open("./models/pointcloud-inpainting.py", "r").read())
+
+##########################################################
+
+
+class Predictor(BasePredictor):
+    def setup(self):
+        """Load the model into memory to make running multiple predictions efficient"""
+        pass
+
+    def predict(
+        self,
+        image: Path = Input(description="Input image"),
+    ) -> Path:
+        """Run a single prediction on the model"""
+
+        npyImage = cv2.imread(filename=str(image), flags=cv2.IMREAD_COLOR)
+        intWidth = npyImage.shape[1]
+        intHeight = npyImage.shape[0]
+
+        fltRatio = float(intWidth) / float(intHeight)
+
+        intWidth = min(int(1024 * fltRatio), 1024)
+        intHeight = min(int(1024 / fltRatio), 1024)
+
+        npyImage = cv2.resize(
+            src=npyImage,
+            dsize=(intWidth, intHeight),
+            fx=0.0,
+            fy=0.0,
+            interpolation=cv2.INTER_AREA,
+        )
+
+        process_load(npyImage, {})
+        objFrom = {
+            "fltCenterU": intWidth / 2.0,
+            "fltCenterV": intHeight / 2.0,
+            "intCropWidth": int(math.floor(0.97 * intWidth)),
+            "intCropHeight": int(math.floor(0.97 * intHeight)),
+        }
+
+        objTo = process_autozoom(
+            {"fltShift": 100.0, "fltZoom": 1.25, "objFrom": objFrom}
+        )
+
+        npyResult = process_kenburns(
+            {
+                "fltSteps": numpy.linspace(0.0, 1.0, 75).tolist(),
+                "objFrom": objFrom,
+                "objTo": objTo,
+                "boolInpaint": True,
+            }
+        )
+
+        output_path = Path(tempfile.mkdtemp()) / "output.mp4"
+        moviepy.editor.ImageSequenceClip(
+            sequence=[
+                npyFrame[:, :, ::-1]
+                for npyFrame in npyResult + list(reversed(npyResult))[1:]
+            ],
+            fps=25,
+        ).write_videofile(str(output_path))
+        return output_path
+
+    # end