Skip to content

farrukh-x2/traffic_sign_classifier_deep_nn

Repository files navigation

Traffic Sign Recognition Classifier

Overview

This project uses deep neural networks and convolutional neural networks to classify traffic signs. A model is trained so it can decode traffic signs from natural images by using the German Traffic Sign Dataset. After the model is trained, it is tested on new images of traffic signs found on the web.

The Project

The goals of this project are achieved following the steps below:

  • Load and visualize the German Traffic Sign Dataset.
  • Preprocess and separate training and validation datasets. Test set is already provided.
  • Design and implement a deep learning model that learns to recognize traffic signs.
  • Model training and tuning of hyperparameters.
  • Determine accuracy of the model and save it for later use.
  • Test model on images from outside the dataset and check confidence of predictions.

Files

  • test_images test images downloaded from the web
  • checkpoint, lenet, lenet.meta are files that contain the saved model
  • signnames.csv a dictionary that assigns labels to the dataset of traffic sign images.

Dataset

Running the first couple of cells will download and unzip the required dataset. The dataset is pickled where the images have already been resized to 32x32.

Usage

Cloning the repository and running the notebook is required to get the project running

Notebook

The deep_nn_traffic_sign_classifier.ipynb notebook goes in thorough details of each of the steps laid out above with accompanying images.

About

Traffic sign classifier based on convolution neural networks

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published