search

YinghaoHuang91 / MuVS

This is the code for MuVS (Multi-view SMPLify)
MIT License
112 stars 23 forks source link

About the Use of Silhouette #7

Open StatML opened 5 years ago

StatML commented 5 years ago

Hi, @YinghaoHuang91 Thanks for your great work! It seems that your codes do not use segmentation results. Can you share that part? Thx!

YinghaoHuang91 commented 5 years ago

Hi I've been busy with something else, and I will try to find some free time to clean and release the segmentation part. If you really want it now, one great source to refer to is the Unite-the-People repository here: https://github.com/classner/up. Thanks!

YinghaoHuang91 commented 5 years ago

Hi, sorry for the way too late response. The idea of silhouette based refinement part is firstly introduced and implemented in Unite-the-People paper (http://files.is.tuebingen.mpg.de/classner/up/), and the shared implementation is dependent on some private code inside our group. It has always been believed that this part of code is already made public in the Unite-the-People repository, however I just recently found that's not really the case. So this functionality will be added after the related code segment is officially released. Thanks again for your attention!

khanhha commented 4 years ago

@YinghaoHuang91 : hi Ying. Thank you for your great work. Could you please share with me the code for the silhouette reprojection error through my email khanhhh89@gmail.com? I find this part very interesting and really want to know the implementation behind the equation. Thank you very much.

YinghaoHuang91 commented 4 years ago

@khanhha Thanks for your interest in our work! I have been planning to add back the silhouette loss term by using most recent Tensorflow based rendering framework. Please keep an eye on this repository.

wine3603 commented 4 years ago

@YinghaoHuang91 hi man, great work! how is the silhouette parts going? Recently I met some problems with python2.7 and opencv-python with numpy versions. Do you have plan to move thie repo to python3, as py2.7 is gonna stop in 5 mouths.

YinghaoHuang91 commented 4 years ago

I've kept being distracted by other ongoing work. I hope I can spare some time working on this around end of this year.

abdullahchand commented 3 years ago

Hi @YinghaoHuang91 , I hope you are doing fine . Any update on adding the sillhouette? and also updating to python 3? . If not can you share a read through which can help me figure out how to add the silloulehte? . Lastly, I did convert your code to python3 but it seems tensorflow was not able to perform optimization at all. After searching a lot I found out that it was an issue of python3. Would be really thankful IF you can help me out

YinghaoHuang91 commented 3 years ago

Hi @abdullahchand I'm working on the new code already, but it's part of another ongoing project, so it won't get released until the new project is done. The new code is based on Python3. For the details of silhouette loss please refer to Equation 1 on Page 3 in this paper: https://openaccess.thecvf.com/content_cvpr_2017/papers/Lassner_Unite_the_People_CVPR_2017_paper.pdf.