search

1adrianb / face-alignment

:fire: 2D and 3D Face alignment library build using pytorch
https://www.adrianbulat.com
BSD 3-Clause "New" or "Revised" License
7.05k stars 1.34k forks source link

What is input and output of depth network? #138

Closed mrgloom closed 5 years ago

mrgloom commented 5 years ago

What is input and output of depth network?

https://github.com/1adrianb/face-alignment/blob/87a496b158ff9a215aa6f48262f4e13d8e6c4dd7/face_alignment/api.py#L49

1adrianb commented 5 years ago

The input is the RGB image stacked alongside the previously produced heatmaps.

mrgloom commented 5 years ago

And what is output? Is it just 68 landmark heatmaps? i.e. it don't predict depthmap? i.e. like https://github.com/AaronJackson/vrn https://github.com/1adrianb/face-alignment/blob/f0cf27cf7c9141f567e135c21495702cdf0aefe3/face_alignment/models.py#L219

1adrianb commented 5 years ago

Aaron's work predict a 3D voxelized volume (DxHxW). This network outputs 68 points, each containing the depth at a given location (x,y).