Features and AU labels not match

haixpham / end2end_AU_speech

Code for the paper "End-to-end Learning for 3D Facial Animation from Speech"

MIT License

69 stars 24 forks source link

Features and AU labels not match #1

Closed yufu closed 5 years ago

yufu commented 5 years ago

Hi haixpham, I have some trouble when reproduce your code. I found my features length are always not equal to the AU labels when using RAVDESS dataset . if it is right ? Waiting for your reply, thanks a lot .

haixpham commented 5 years ago

Hi zuglerfu, No, there are not that many mismatched input-label sequences. The reason is simply because the tracker may miss one or two frames of the video when the actor turns their face away from camera. You can run create_spectrogram_CTF.py to see the text message, there are a couple of dozens of those mismatched sequences at most.