andrewjong/SwapNet

Implements garment transfer by decomposing images into texture and body segmentation channels, then using a deep generative model to synthesize clothing swaps while preserving body structure. The pipeline integrates Neural Body Fitting for pose-aware body segmentation and LIP_JPPNet for cloth parsing, with preprocessing utilities and a Visdom visualization server for training monitoring. Supports both containerized Docker deployment and conda-based local installation on Linux.

390 stars. No commits in the last 6 months.