Implemented a Convolutional Neural Network to detect traffic signs in a full stack web application
- Recognizes 43 classes of traffic signs from the Kaggle Traffic Sign data set
- Validation set accuracy of 99.2% and test set accuracy of 96.6%
- Reduced overfitting with regularization techniques such as dropout layers, max pooling, and a validation set
- Varied model parameters including batch size, dropout rate, and activation functions to improve accuracy
App Demo: