Run author's implementation

vikasTmz / ufc

Implementation of "Unrestricted Facial Geometry Reconstruction Using Image-to-Image Translation" by Sela et al.

2 stars 0 forks source link

Open vikasTmz opened 4 years ago

vikasTmz commented 4 years ago

vikasTmz commented 4 years ago

Steps in the Authors Code

Extracts and crops the face regions using Matlab's vision.CascadeObjectDetector.
Resize image to 512x512 (because this is the input dimension to the image-to-image network).
Run the torch models for depth and correspondence estimation. See this issue