XingangPan/GAN2Shape
Code for GAN2Shape (ICLR2021 oral)
Reconstructs 3D object shapes and geometry from 2D StyleGAN-generated images through unsupervised learning, leveraging the implicit 3D knowledge encoded in pre-trained GANs without manual annotations. Uses neural rendering with differentiable mesh projection to optimize latent codes and camera parameters, enabling downstream tasks like object rotation and relighting. Built on PyTorch and neural_renderer, with support for multiple object categories (faces, cars, churches) and joint pre-training pipelines for improved results.
576 stars. No commits in the last 6 months.
Stars
576
Forks
99
Language
Python
License
MIT
Category
Last pushed
Jul 04, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/XingangPan/GAN2Shape"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
rundiwu/DeepCAD
code for our ICCV 2021 paper "DeepCAD: A Deep Generative Network for Computer-Aided Design Models"
ayaanzhaque/instruct-nerf2nerf
Instruct-NeRF2NeRF: Editing 3D Scenes with Instructions (ICCV 2023)
compphoto/Intrinsic
Repo for the papers "Intrinsic Image Decomposition via Ordinal Shading" (TOG 2023) and "Colorful...
mworchel/differentiable-shadow-mapping
Differentiable Shadow Mapping for Efficient Inverse Graphics (CVPR 2023)
NVlabs/dream-in-4d
Official PyTorch implementation of "A Unified Approach for Text- and Image-guided 4D Scene...