- Train a model on German Traffic Sign Recognition Benchmark Dataset.
- Introduce 5 new classes to the dataset.
- Evaluate results of Benchmark network on this new dataset.
- Analyse and visualize various metrics of the network on the new dataset.
- Recommend steps to be taken to improve the model performance.
- Make changes to the network/dataset based on above suggestions.
- Make a UI for easy addition of classes and images to various classes of existing dataset. A platform where user can easily add images, augment the dataset and segregate the dataset.
- Connect the output of classifier to UI so that user can visualize the performance on the UI and give some suggestions on the UI for network performance improvement.
Baseline Model on GTSRB - 98.60 % Baseline Model on New Dataset - 96.37%
Baseline Model in GTSRB - 96.82% Baseline Model on New Dataset - 94.90%
