The shuffle op implement seems like time consuming, does anyone has better idea？

610265158 / face_landmark

A simple method for face alignment based on wingloss and mutitask learning :)

Apache License 2.0

251 stars 80 forks source link

Closed 610265158 closed 4 years ago