- First Learn about CNNs at MNIST-CNN.ipynb
- Then learn about Transfer learning from CVFineTuning.ipynb
- Implement a CNN from scratch on cats and dogs using 2D convolutions and max pooling. For bonus points use Data Augmentation. How do we do?
- Transfer learn from vgg16 instead of xception on cats and dogs. How do the results progress through creating features, unfreezing only the head, unfreezing the last layers of vgg16, to unfreezing all of vgg16?
We'll discuss it on the #code-with-rahul
channel at our Discord.