search

ZitongYu / STVEN_rPPGNet

Main code of "Remote Heart Rate Measurement from Highly Compressed Facial Videos: an End-to-end Deep Learning Solution with Video Enhancement"
MIT License
155 stars 44 forks source link

what is the input label for target label mask? #9

Closed alphapibeta closed 2 years ago

alphapibeta commented 3 years ago

Step 2: Forward model and calculate the losses input 1: facial frames --> [3, 64, 128, 128]

input 2: target label mask --> 5D vector

video_GroudTruth: the original video (before highly compressed)

•Can someone explain to me, how the target label mask is 5D? •How do I prepare the dataset?

wang-yu1016 commented 3 years ago

how to make binary skin labels