search

lmb-freiburg / hand3d

Network estimating 3D Handpose from single color images
GNU General Public License v2.0
802 stars 252 forks source link

Forward pass? #17

Closed jayala closed 5 years ago

jayala commented 5 years ago

Hello, what does the forward pass weights contains? Was it trained with the whole RHD dataset for handsegnet and posenet? Because it says minimal example should I download the dataset and retrain all?

zimmerm commented 5 years ago

Hi, for running the minimal example you dont need to download the dataset. Only download the network weights.

El dom., 7 oct. 2018 a las 20:15, Jesús Ayala (notifications@github.com) escribió:

Hello, what does the forward pass weights contains? Was it trained with the whole RHD dataset for handsegnet and posenet? Because it says minimal example should I download the dataset and retrain all?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lmb-freiburg/hand3d/issues/17, or mute the thread https://github.com/notifications/unsubscribe-auth/ASwP_4L2ZXLaNkHIz9rzpZqIFaxD5OFLks5uisMlgaJpZM4XMMdn .

jayala commented 5 years ago

Oh, yeah I did that. Got interesting results with the forward pass, but somehow it sometimes confuses lips with hands.

https://image.ibb.co/cz5OD0/outoutx-out.png

Do you think this there is something wrong with my setup?

zimmerm commented 5 years ago

No not necessarily. Truthfully, It is not the strongest pose estimator out there anymore and the behavior you describes partially stems from its assumption that there is always a hand present in the image.

jayala commented 5 years ago

Yeah, that's what I thought too. Well, will try some of the other projects out there. Thank you @zimmerm