search

Mark12Ding / FAME

Code for Motion-aware Contrastive Video Representation Learning via Foreground-background Merging (CVPR 2022)
https://arxiv.org/abs/2109.15130
Apache License 2.0
45 stars 5 forks source link

Can you provide the pretraining log of UCF101? #5

Open Randle-Github opened 2 years ago

Randle-Github commented 2 years ago

Hi, I confronted a problem with your code but transfer to some other datasets that the contrastive loss could hardly drop. I wonder if it is a normal phenomenon that such loss could not demonstrate the results of the model or some other problems. If possible, can you provide your log with UCF101 or just provide some details of the loss curve? Thanks.

Mark12Ding commented 2 years ago

Well, the loss does drop slightly at the pretraining stage due to the harder objective in our setting. And the loss curve doesn't indicate whether the model is pre-trained well or not. Instead, the results of the downstream tasks can prove it.