MODNet
MODNet copied to clipboard
The outcome is different from Collab and the demo website?
Trying with the same image on Google Collab and https://sight-x.cn/portrait_matting/ But the outcome is totally different, how to fix this bug?
I have this problem too. Are the trained models different or some preprocessing? please help to solve this issue.
@maplerichie @zara Hi, Both, thanks for your attention. As said in the README file. The models used in the website and the model released in this Github are different. Therefore, you got inconsistent results.
@maplerichie @zara Hi, Both, thanks for your attention. As said in the README file. The models used in the website and the model released in this Github are different. Therefore, you got inconsistent results.
Thanks for your response, do you have any schedule to publish the advance model?
Also very looking forward to the more advanced model !!! The performance of the online demo is very impressive, great work!!!
@zara I am sorry that I cannot publish the model in the online demo at this moment, but we expect to open source more powerful models to the community in the future.
What is the difference between the model in collab vs the website model? Can you briefly give comparison?
When do you intend to publish the advanced model?