search

YoungSeng / QPGesture

QPGesture: Quantization-Based and Phase-Guided Motion Matching for Natural Speech-Driven Gesture Generation (CVPR 2023 Highlight)
82 stars 7 forks source link

About the fps #17

Closed huifu99 closed 1 year ago

huifu99 commented 1 year ago

You mentioned you used 60 fps for all methods in your paper, but I found that you align all poses with 20 fps in your code 'make_beat_dataset.py'. image image

YoungSeng commented 1 year ago

Yes, our model saves data at low sampling rates but doesn't use it because all baseline models are implemented at low sampling rates, e.g. CaMN at 15fps, KNN at 15fps, End2End at 20fps...

See make_beat_dataset.py https://github.com/YoungSeng/QPGesture/blob/ef0f2a93f469d7e227896f901c8b1ddda8b2100a/process/make_beat_dataset.py#L162

https://github.com/YoungSeng/QPGesture/blob/ef0f2a93f469d7e227896f901c8b1ddda8b2100a/process/make_beat_dataset.py#L182

https://github.com/YoungSeng/QPGesture/blob/ef0f2a93f469d7e227896f901c8b1ddda8b2100a/process/make_beat_dataset.py#L161

https://github.com/YoungSeng/QPGesture/blob/ef0f2a93f469d7e227896f901c8b1ddda8b2100a/process/make_beat_dataset.py#L188