[NeurIPS 2022 Spotlight] VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
1.32k
stars
131
forks
source link
How good model handle the different duration of clips? #116
Closed
hagonata closed 2 months ago
It's okay if I will have 10 seconds with 2-5 seconds duration clips in one class?