HongwenZhang / JVCR-3Dlandmark

[TIP 2019] Adversarial Learning Semantic Volume for 2D/3D Face Shape Regression in the Wild
MIT License
152 stars 30 forks source link

The issue for generating 3D heatmaps #3

Open lx120 opened 6 years ago

lx120 commented 6 years ago

Hi, Zhang

Thanks for your great work.

I try to generate ground truth volumes representation about 300W-LP based on your code utils.transforms.creat_volume and utils.imutils.draw_labelvolume.

But it does not same as your volumes representation generated from the net model you have trained, could you tell me how to generate correct ground truth volumes representation for 300W-LP.

Thanks!

HongwenZhang commented 6 years ago

Duplicate of #1