search

MCG-NJU / VideoMAE

[NeurIPS 2022 Spotlight] VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
https://arxiv.org/abs/2203.12602
Other
1.32k stars 131 forks source link

VIT-S initilization #80

Closed G-JWLee closed 1 year ago

G-JWLee commented 1 year ago

If I searched correctly, image mae does not provide Vit-S version. How did you initilized model weight when training on Vit-S?

yztongzhan commented 1 year ago

All our models are pre-trained from scratch on video datasets.