I tuned different Pytorch/Tensorflow pre trained models like ResNet50 , Wide ResNet_50.2 , VGG16 and a custom CNN model to classify a dog image among 120 breeds. The training dataset can be collected from here.

Test Accuracy : 82.88978494623656
Test Loss : 0.6641809101546964

• Loss :

  

• Accuracy :

  

Test Accuracy : 80.26881720430107 / 0.9375 (Tensorflow) Test Loss : 0.6438697055783323

• Loss :

  

• Accuracy :

  

Test Accuracy : 80.8736559139785
Test Loss : 0.6329619951385964

• Loss :

  

• Accuracy :

  

Test Accuracy : 0.1548913043478262 / 0.6875 (Tensorflow) Test Loss : 4.78740421585414

• Loss :

  

• Accuracy :