Projects/Cats vs Dogs

This is a classification project that classifies whether the give image is a cat or a dog. The dataset is taken from the Kaggle. Dataset composition: 1. training set = 20000 2. validation set = 3000 3. validation set = 2000

I've used different approches to train the model. A simple convolution network with 3 layers, 1 fully connected layer and a output layer which gives accuracy of 89.97% on validation set and 88.70% on test set. Then i trained another model with pre-trained architecture (MobileNet) which gives accuracy of 97.76 on validation set and 97.95% on test set.

#GRAPH

Convolutional Network vs Pre-trained Architecture (MobileNet)

COVNET loss: ------------------------------------------------------------------------------ MobileNet loss:

COVNET accuracy --------------------------------------------------------------------------- MobileNet accuracy