This is project 2 in course Deep learning Nano Degree by udacity.
Download the dog dataset. https://s3-us-west-1.amazonaws.com/udacity-aind/dog-project/dogImages.zip
Download the human dataset. https://s3-us-west-1.amazonaws.com/udacity-aind/dog-project/lfw.zip
Here we implement CNN to predict dog breed and at the end make a function that will detect between human and a dog and show top 3 prediction.
In first part, We build CNN from Scratch with pytorch. I have tried to imitate the VGG16 model.(accuracy 12%)
In the second part. We used VGG16 trained model for transfer learning to predict the dog breeds.(accuracy 75%)