Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
[CVPR '21] Official repository for Few-shot Image Generation via Cross-domain Correspondence
| Date | Stars |
|---|---|
| 2026-07-24 | 304 |
| 2026-07-25 | 304 |
| 2026-07-28 | 304 |
| 2026-07-30 | 304 |
| 2026-08-06 | 304 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# Few-shot Image Generation via Cross-domain Correspondence [Utkarsh Ojha](https://utkarshojha.github.io/), [Yijun Li](https://yijunmaverick.github.io/), [Jingwan Lu](https://research.adobe.com/person/jingwan-lu/), [Alexei A. Efros](https://people.eecs.berkeley.edu/~efros/), [Yong Jae Lee](https://web.cs.ucdavis.edu/~yjlee/), [Eli Shechtman](https://research.adobe.com/person/eli-shechtman/), [Richard Zhang](https://richzhang.github.io/) Adobe Research, UC Davis, UC Berkeley  PyTorch implementation of adapting a source GAN (trained on a large dataset) to a target domain using very few images. ### [Project page](https://utkarshojha.github.io/few-shot-gan-adaptation/) | [Paper](https://arxiv.org/abs/2104.06820) ## Overview <img src='imgs/method_diagram.png' width="840px"/> Our method helps adapt the source GAN where one-to-one correspondence is preserved between the source G<sub>s</sub>(z) and target G<sub>t</sub>(z) images. ## Requirements **Note:** The base model is taken from [StyleGAN2](https://github.com/rosinality/stylegan2-pytorch)'s implementation from [@rosinality](https://github.com/rosinality) - Linux - NVIDIA GPU + CUDA CuDNN 10.2 - PyTorch 1.7.0 - Python 3.6.9 - Install all the libraries through `pip install -r requirements.txt` ## Testing We provide the pre-trained models for different source and adapted (target) GAN models. | Source GAN: G<sub>s</sub> | Target GAN: G<sub>s→t</sub> | | ------------------------- | --------------------------------- | | [FFHQ](https://drive.google.com/file/d/1TQ_6x74RPQf03mSjtqUijM4MZEMyn7HI/view?usp=sharing) | [[Sketches](https://drive.google.com/file/d/1Qkdeyk_-1pqgvrIFy6AzsSYNgZMtwKX3/view?usp=sharing)] [[Caricatures](https://drive.google.com/file/d/1CX8uYEWqlZaY7or_iuLp3ZFBcsOOXMt8/view?usp=sharing)] [[Amedeo Modigliani](https://drive.google.com/file/d/1WvBtThEakKNqNFBCuHHoNNI1jojFAvan/view?usp=sharing)] [[Babies](https://drive.google.com/file/d/1d5JNwQBSyFaruAoLZBlXFVPc_I6WZjhm/view?usp=sharing)] [[Sunglasses](https://drive.google.com/file/d/1D6HOdcHG4j6kQmOCjwQakK7REgykPOYy/view?usp=sharing)] [[Rafael](https://drive.google.com/file/d/1K6xWnlfQ-qT_I_QTY8SiQ9fvRylMFeND/view?usp=sharing)] [[Otto Dix](https://drive.google.com/file/d/1I8gmuiDcARmwZNimlYEalPsKcRot-ijZ/view?usp=sharing)] | | [LSUN Church](https://drive.google.com/file/d/18NlBBI8a61aGBHA1Tr06DQYlf-DRrBOH/view?usp=sharing) | [[Haunted houses]()] [[Van Gogh houses]() [[Landscapes]()] [[Caricatures]()] | | [LSUN Cars](https://drive.google.com/file/d/1O-yWYNvuMmirN8Q0Z4meYoSDtBfJEjGc/view?usp=sharing) | [[Wrecked cars]()] [[Landscapes]()] [[Haunted houses]()] [[Caricatures]()] | | [LSUN Horses](https://drive.google.com/file/d/1ED4JPQsxnBUMFHiooCL7oK2x4FfIf-bt/view?usp=sharing) | [[Landscapes]()] [[Caricatures]()] [[Haunted houses]()] | | [Hand gestures](https://drive.google.com/file/d/1LBXphEMT8C2JJ3AXf2CETeIFvoAz5n2T/view?usp=sharing) | [[Google Maps]()] [[Landscapes]()] | For now, we have only included the pre-trained models using FFHQ as the source domain, i.e. all the models in the first row. We will add the remaining ones soon. Download the pre-trained model(s), and store it into `./checkpoints` directory. ### Sample images from a model To generate images from a pre-trained GAN, run the following command: ```bash CUDA_VISIBLE_DEVICES=0 python generate.py --ckpt_target /path/to/model/ ``` Here, `model_name` follows the notation of `source_target`, e.g. `ffhq_sketches`. Use the `--load_noise` option to use the noise vectors used for some figures in the paper (Figures 1-4). For example: ```bash CUDA_VISIBLE_DEVICES=0 python generate.py --ckpt_target ./checkpoints/ffhq_sketches.pt --load_noise noise.pt ``` This will save the images in the `test_samples/` directory. ### Visualizing correspondence results To visualize the same noise in the source and adapted models, i.e. G<sub>s</sub>(z) and G<s
Excerpt of 13,463 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:e0ae2560666d2f29, topic:computer-vision