The size of input video tensor for X3D

facebookresearch / SlowFast

PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models.

Apache License 2.0

6.5k stars 1.2k forks source link

The size of input video tensor for X3D #307

Open KevinQian97 opened 3 years ago

KevinQian97 commented 3 years ago

Hi, I am curious about the size of the input tensor for X3D. Usually, it should be Batch Channel T H W. However, in your implementation, there is another parameter called "pathway", So, BCTHW apparently does not work here. By the way, would you mind telling me the mean and std for the normalization transform of input videos? I didn't find that in your code. Thanks!

Shiro-LK commented 3 years ago

@KevinQian97 Hi , I am also interested by this information. Did you succeed to find out ?

ZainZhao commented 3 years ago

I also want to know what is that ?

thearkamitra commented 10 months ago

If the model is a single_pathway model, you can just put it in a list and you are good to go. If the model has multiple pathways, you need to create another input of "slower" length and have a list [slow, fast].