udacity/deep-learning-v2-pytorch

Covers CNN architectures (with layer visualization), RNNs for sequence modeling, GANs including CycleGAN for unpaired image translation, and transfer learning with pre-trained models like VGGNet. Built entirely in PyTorch with Jupyter notebooks, the material progresses from foundational gradient descent through advanced topics like attention mechanisms, style transfer, and AWS SageMaker deployment. Includes hands-on projects spanning computer vision (dog breed classification, face generation) and NLP (sentiment analysis, TV script generation with character-level RNNs).

5,486 stars. No commits in the last 6 months.