search

MartinXM / GAP

official implementation for Language Supervised Training for Skeleton-based Action Recognition
Apache License 2.0
96 stars 11 forks source link

About NW-UCLA #4

Open SunPengP opened 1 year ago

SunPengP commented 1 year ago

Hello, when i tried to use CTR-GCN to train NW-UCLA, i got 95.5 accurancy , 1% lower than the result CTR-GCN given. Did you meet the same problem in your experiments.

MartinXM commented 1 year ago

You may need to tune the weights of different modals.

SunPengP commented 1 year ago

Hello, when i tried to use CTR-GCN to train NW-UCLA, i got 95.5 accurancy , 1% lower than the result CTR-GCN given. Did you meet the same problem in your experiments.

You may need to tune the weights of different modals.

Thank you for your reply. Is this how you handled your experiment