======= Explore CIFAR-10 DATASET
-
Used VGG16 model ,adding some dense connected layer at top,and using finetunning and transferlearning by which i got an overfit model with 96% train accuracy ,And 82.41% test accuracy
-
Used a model same as VGG16 structure but with dropout,Which decrease the overfitting. Achieved a train accuracy of 87% and test accuracy of 84%.Traning for more epochs may help out as the model is getting underfit
-
Using Dataaugmentaion with same model used in second and achieved a trainaccuracy of 86% and test accuracy of 89%. It seems quite good model,Traning for more epoches on changing costfunction and all can increase the accuracy.