wpeebles/gangealing
A PyTorch implementation of a CVPR 2022 paper that trains a spatial transformer to align GAN-generated images to a learned canonical mode.
Not currently ranked — collecting fresh signals.
star history
GANgealing applies GAN-supervised learning to dense visual alignment, training a spatial transformer to warp unaligned GAN samples to a common target mode. The method jointly learns both the target mode and the alignment transform end-to-end. The model generalizes to real images after training exclusively on synthesized GAN data.
Frequently asked
- What is wpeebles/gangealing?
- A PyTorch implementation of a CVPR 2022 paper that trains a spatial transformer to align GAN-generated images to a learned canonical mode.
- Is gangealing open source?
- Yes — wpeebles/gangealing is open source, released under the BSD-2-Clause license.
- What language is gangealing written in?
- wpeebles/gangealing is primarily written in Python.
- How popular is gangealing?
- wpeebles/gangealing has 1k stars on GitHub.
- Where can I find gangealing?
- wpeebles/gangealing is on GitHub at https://github.com/wpeebles/gangealing.