Paper-Implementations

Why we included this project

This is a small, readable collection of PyTorch reimplementations of classic deep learning papers, useful as a study aid rather than something you would deploy. Each folder turns a well-known idea into working code: CNN variants like Network in Network and ResNet18 trained on MNIST, generative models including DCGAN, pix2pix, CycleGAN, and VAE, several neural style transfer approaches, and a character-level RNN. For students or engineers who want to see how the math in a paper becomes actual tensors and training loops, having these examples side by side makes it easy to compare architectures and spot the differences between them. The code is deliberately small and self-contained, so it is a convenient starting point to experiment with or adapt into a minimal reference implementation. It is not a maintained framework, but as a companion for studying classic computer vision and generative modeling it still holds up.