3D Garment extraction using cloth landmarks
Worked on training HRNet using DeepFasion 2D dataset to estimate cloth landmarks in reference images. Used these estimated key points and 3D reconstructed humans from a single view to perform garment parsing.
Topic Covered:
- Orthographic projection models and coordinate systems.
- Multi-GPU training - Data-Parallel and Distributed Data-Parallel.
Programming Language: Pytorch, Python