Caltech101

Link to dataset: http://www.vision.caltech.edu/Image_Datasets/Caltech101/

I have used this dataset to perform image classification task.

I used Transfer Learning for solving the problem.The used architecture was VGG16. It acheieved the accuracy of 79.34% with least diversion in train accuracy and validation accuracy; i.e the model is not overfitting. Data augmentation helped me to acheiece least diversion in train accuracy and validation accuracy.

Again I used InceptionV3 architecture and boom the accuracy on unseen data reached to 90.39%; but when I used data augmentation it dropped to 88.49%.