In these notebooks we build convolutional neural networks with TensorFlow/Keras. One notebook does it from "scratch" and the other utilizes Transfer Learning with the pretrained Xception network.
The accuracy difference between both approaches is huge:
| Notebook | Accuracy |
|---|---|
| DogBreed Classifier.ipynb | 37 % |
| DogBreed Classifier TransferLearning.ipynb | 94 % |
The dataset, called , is a subset of dog pictures from ImageNet and consists of 10,000 training and 3,000 test data. It contains the dog breeds
- Australian Terrier
- Border Terrier
- Samoyed
- Beagle
- Shih-Tzu
- English Foxhound
- Rhodesian Ridgeback
- Dingo
- Golden Retriever
- Old English Sheepdog.