Image-classifier

This project involves building an image classifier using a custom model that resembles ResNet 34 and also employing transfer learning using a pretrained MobileNet network. Learning rate range test is employed to determine the apt learning rate. Data augmentaion is used to regularize the model. The models are trained with a train set of size 4000 images and validated on a validation set of size 1000 images. I have also tried building a classifier using SVM trained with SIFT features in bag of visual words representation.