This small project was created as a part of the CS50ai 2024 course. The CNN was trained on the GTSRB dataset with the following results on the final model.
The training and testing metrics indicate strong model performance, with training accuracy of 93.63% and testing accuracy of 96.83%. The precision and recall values are also high, meaning that the model has good generalization and is effectively handling both false positives and false negatives.
training:
- accuracy: 0.9363 - loss: 0.1927 - precision: 0.9535 - recall: 0.9233
testing:
- accuracy: 0.9683 - loss: 0.1042 - precision: 0.9755 - recall: 0.9632