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.
Sketch Your Own GAN: Customizing a GAN model with hand-drawn sketches.
| Date | Stars |
|---|---|
| 2026-07-24 | 708 |
| 2026-07-25 | 708 |
| 2026-07-28 | 708 |
| 2026-07-30 | 708 |
| 2026-08-06 | 708 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
## Sketch Your Own GAN
[**Project**](https://peterwang512.github.io/GANSketching/) | [**Paper**](https://arxiv.org/abs/2108.02774) | [**Youtube**](https://www.youtube.com/watch?v=1smwQ-tSCjM) | [**Slides**](https://drive.google.com/file/d/1ADtaOLauX3h-mB3tmjdMFs1hcnJNKZwO/edit)
<img src="images/teaser_video.gif" width="800px"/>
Our method takes in one or a few hand-drawn sketches and customizes an off-the-shelf GAN to match the input sketch. While our new model changes an object’s shape and pose, other visual cues such as color, texture, background, are faithfully preserved after the modification.
<br><br><br>
[Sheng-Yu Wang](https://peterwang512.github.io/)<sup>1</sup>, [David Bau](https://people.csail.mit.edu/davidbau/home/)<sup>2</sup>, [Jun-Yan Zhu](https://cs.cmu.edu/~junyanz)<sup>1</sup>.
<br> CMU<sup>1</sup>, MIT CSAIL<sup>2</sup>
<br>In [ICCV](https://arxiv.org/abs/2108.02774), 2021.
**Aug 16 Update** Training code, evaluation code, and dataset are released. Model weights are also updated, please re-run `bash weights/download_weights.sh` if you have downloaded the weights before this update.
## Results
Our method can customize a pre-trained GAN to match input sketches.
<img src="images/teaser.jpg" width="800px"/>
**Interpolation using our customized models.** Latent space interpolation is smooth with our customized models.
<table cellpadding="0" cellspacing="0" >
<tr>
<td align="center">Image 1 <br> <img src="images/cat2.jpg" width=240px></td>
<td align="center">Interoplation <br> <img src="images/interp2.gif" width=240px></td>
<td align="center">Image 2 <br> <img src="images/cat1.jpg" width=240px></td>
</tr>
</table>
**Image editing using our customized models.** Given a real image (a), we project it to the original model's latent space z using [Huh et al.](https://github.com/minyoungg/pix2latent) (b). (c) We then feed the projected z to the our standing cat model trained on sketches. (d) Finally, we showed edit the image with `add fur` operation using [GANSpace](https://github.com/harskish/ganspace).
<img src="images/editing.jpg" width="800px"/>
**Model interpolation.** We can interpolate between the customized model by interpolating the W-latent space.
<table cellpadding="0" cellspacing="0" >
<tr>
<center><td align="center">Model 1 <br> <img src="images/model_interp/1.png" height=90px border=1></td></center>
<center><td align="center">Interoplation in W-latent space <br> <img src="images/model_interp/latent_1.gif" height=90px></td></center>
<center><td align="center">Model 2 <br> <img src="images/model_interp/0.png" height=90px border=1></td></center>
</tr>
</table>
<table cellpadding="0" cellspacing="0" >
<tr>
<center><td align="center"><img src="images/model_interp/1.png" height=90px border=1></td></center>
<center><td align="center"><img src="images/model_interp/latent_2.gif" height=90px></td></center>
<center><td align="center"><img src="images/model_interp/2.png" height=90px border=1></td></center>
</tr>
</table>
We observe similar effect by interpolating the model weights directly.
<table cellpadding="0" cellspacing="0" >
<tr>
<center><td align="center">Model 1 <br> <img src="images/model_interp/1.png" height=90px border=1></td></center>
<center><td align="center">Interoplation in the model weight space <br> <img src="images/model_interp/weight_1.gif" height=90px></td></center>
<center><td align="center">Model 2 <br> <img src="images/model_interp/0.png" height=90px border=1></td></center>
</tr>
</table>
<table cellpadding="0" cellspacing="0" >
<tr>
<center><td align="center"><img src="images/model_interp/1.png" height=90px border=1></td></center>
<center><td align="center"><img src="images/model_interp/weight_2.gif" height=90px></td></center>
<center><td align="center"><img src="images/model_interp/2.png" height=90px border=1></td></center>
</tr>
</table>
**Failure case**. Our method is nExcerpt of 10,287 characters
Read on GitHub
Jun-Yan Zhu · Carnegie Mellon University: · United States
10
10
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:7e9e441e5ecb660c, topic:deep-learning
matched fp:7e9e441e5ecb660c, topic:computer-vision