Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set (CVPRW 2019). A PyTorch implementation.
Role in this project:
ML Engineer Contributions:17 commits, 8 PRs, 21 pushes in 1 year 2 months
Contributions summary:Sicheng primarily contributed to the development and refinement of the 3D face reconstruction model within the PyTorch framework. Their work involved adjusting loss function weights, fixing bugs related to mesh exporting, and adding new features like an OpenGL rendering option. The commits demonstrate an understanding of the model's architecture and the differentiable rendering pipeline. The contributions directly impact model functionality and control.
3dimagesetpytorchweakly-supervised-learning
Deep 3D Portrait from a Single Image
Contributions:11 commits, 10 pushes, 1 branch in 6 months
3d