This is a project done with regard to the Aritifical Intelligence module at SRH. Classification of German Traffic Signs

This project uses Multi-Class CNN to classify images.

The dataset used is from Kaggle: https://www.kaggle.com/meowmeowmeowmeowmeow/gtsrb-german-traffic-sign

Data Augmentation Dropout Hidden Layers Optimizer Learning Rate Epochs Training Set Accuracy Validation Set Accuracy Test Set Accuracy No Yes 1-256 RMSPROP 0.001 20 0.974751353263855 0.949496209621429 0.951148060174188 Yes Yes 1-256 RMSPROP 0.001 20 0.959831655025482 0.952174484729767 0.956215360253365

No No Dropouts used 1-256 RMSPROP 0.001 20 0.997067093849182 0.960336685180664 0.953602533650039 Yes No Dropouts used 1-256 RMSPROP 0.001 20 0.993560314178467 0.949496209621429 0.954631828978622

No Yes 1-256,2-128 RMSPROP 0.001 6 0.972551643848419 0.959061324596405 0.94061757719715 Yes Yes 1-256,2-128 RMSPROP 0.001 7 0.956101775169373 0.928453028202057 0.942676167854315

No Yes 1-256,2-128 RMSPROP 0.001 20 0.958524584770202 0.954342544078827 0.951148060174188 Yes Yes 1-256,2-128 RMSPROP 0.001 20 0.94102269411087 0.937890589237213 0.957482185273159----- Best Accuracy for Test SET

No No Dropouts used 1-256,2-128 RMSPROP 0.001 20 0.996238231658936 0.957148313522339 0.947980997624703 Yes No Dropouts used 1-256,2-128 RMSPROP 0.001 20 0.991264998912811 0.940441250801086 0.940380047505938

No Yes 1-256,2-128,3-64 RMSPROP 0.001 20 0.925146639347076 0.929728329181671 0.93396674584323 Yes Yes 1-256,2-128,3-64 RMSPROP 0.001 20 0.859920918941498 0.909322798252106 0.922486144101346

No No Dropouts used 1-256,2-128,3-64 RMSPROP 0.001 20 0.994197905063629 0.948858559131622 0.941567695961995 Yes No Dropouts used 1-256,2-128,3-64 RMSPROP 0.001 20 0.983837008476257 0.919780611991882 0.937925574030087