Open LiuXiaolong19920720 opened 3 years ago
Great job! Which dataset are you using for human image matting?
Great job! Which dataset are you using for human image matting?
I use this human matting dataset and a custom dataset.
Great job Xiaolong. Thank you very much for exploring the inspiring and awesome applications of u2net. We are also trying similar tasks. Hope we can have change to collaborate later. Best of luck.
After a new more than 10 days training, I deployed a head matting model on my website: CVPy-Head Matting.
@LiuXiaolong19920720 is that model available?
Hi, Xiaolong,
The results look awesome. Would you mind that I add your webpage or the corresponding github repo in the readme of U^2-Net ?
On Sun, Dec 13, 2020 at 2:43 AM LiuXiaolong notifications@github.com wrote:
After a new more than 10 days training, I deployed a head matting model on my website: CVPy-Head Matting https://www.cvpy.net/studio/cv/func/DeepLearning/matting/head_matting/page/ .
[image: head matting] https://camo.githubusercontent.com/1189e03a024f86a8f56eef6d65ca130f467a8c424cc1eaa8f6943daff6ee25fc/68747470733a2f2f696d672e637670792e6e65742f686561645f6d617474696e675f6578616d706c655f312e706e67
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NathanUA/U-2-Net/issues/111#issuecomment-743980066, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGOROUXY6FTH3Z7455RPLSUSEDFANCNFSM4UIZTV2Q .
-- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/
Hi, Xiaolong, The results look awesome. Would you mind that I add your webpage or the corresponding github repo in the readme of U^2-Net ?
It's a great honor if you could add the webpage in the readme of U^2-Net.
@LiuXiaolong19920720 is that model available?
It is not available cause it was trained on a private custom dataset with more than 10k portrait images.
Hi, Xiaolong, I just added your webpage link in the ReadMe file of U^2-Net so that others can find your webpage more easily. BTW, if you want, you can also provide me some of your visual results just like what you showed in this issue (if you agree, I can use these visual results), I can then also include these visual results in the ReadMe file and that will attract more attentions I think.
Hi, Xiaolong, I just added your webpage link in the ReadMe file of U^2-Net so that others can find your webpage more easily. BTW, if you want, you can also provide me some of your visual results just like what you showed in this issue (if you agree, I can use these visual results), I can then also include these visual results in the ReadMe file and that will attract more attentions I think.
Thank you for adding the webpage to the ReadMe file. I will collect some visual results and send them to you.
Glad to see the latest update about the human portrait drawing. @NathanUA said they found an interesting application of U^2-Net for human portrait drawing. The application they found should be the application that I developed two month ago based on the amazing U^2-Net and APDrawingGAN dataset, I deployed the model here: CVPy-Human portrait drawing。Welcome to try!
After training on APDrawingGAN dataset, I generated another dataset by the first trained model on a face dataset and train again, so the final model looks more smooth.
At the same time, I also deployed the author's original model here: CVPy-Image Matting. Yes, I use the original model as a matting model... And there are a lot of interesting ways to play, such as matting and sketching.![](https://img.cvpy.net/matting_img2sketch_example_202011011537.jpg)
Recently, I am modifying and retraining the model to solve the problem of imperfect edge matting such as hair region when human image matting. After 5 days training, some test results of one of intermediate models are as follows:
![](https://img.cvpy.net/stage_4_1.png)
This model hasn't been deployed yet because I'm still optimizing it.
Thanks to the author for his wonderful work! Welcome to exchange.