bipinkc19 / image-classifier-cat-dog

Image Classifier for Cat or Dog

• Used an simple CNN to implment the technique in which acheived around 74% accuracy using 8000 test and 2000 train images.
• Used transfer learning from inceptionv3 and adding a single dense layer at the end of it.
• Acheives 100% acuuracy from transfer learning as inceptionv3 is already trained for cat or dog.

Used tensorflow to train the model.

Created beautiful vizualization of graphs and accuracy growth in tensorboard.

Used Beholder to see which nodes are being trained and which are useless.

Used pickle to load and train imgaes after augmentation in Keras.