Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

trained for 3 days but results are very strange #44

Open
hassanrazakhalid opened this issue Apr 17, 2020 · 1 comment
Open

trained for 3 days but results are very strange #44

hassanrazakhalid opened this issue Apr 17, 2020 · 1 comment

Comments

@hassanrazakhalid
Copy link

hi
i trained on company dataset for almost 3-4 days on GTX 1080TI. Please see the log file. Error was keep getting low as expected but when i tested even on trained images. It is behaving strangely
log.txt

Please see the orignal image and result after training. Expect for a few white dashes nothing is there.
out
1803151818-00000004

Any idea what i did wrong here. Note i generated alphas by seprating the alpha channel in provided masks and not KNN. Sample orignal GT is
1803151818-00000008

@Plwy
Copy link

Plwy commented Jul 22, 2020

if you don't have ground truth alpha matte, you need to generate. The mask is just a coarse seg of the person ,can not represent the ground truth alpha matte .you need to use the mask to generate the trimap ,then use trimap and origin image as the input of knn-matting to generate alpha matte.We use the alpha matte that knn-matting generate as the approximation of ground truth alpha matte.

# for free to join this conversation on GitHub. Already have an account? # to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants