mmaction2 icon indicating copy to clipboard operation
mmaction2 copied to clipboard

Published 20 hours ago • verified publisher icon open-mmlab

ImageNet-1K models

Open owaisCS opened this issue 2 years ago • 3 comments

I noticed that in this Github repo you have uploaded the TimeSformer model pre-trained on ImageNet-21K, Could you please update/send the TimeSformer model pre-trained on ImageNet-1K? Or else as mentioned in your repo in point 3, image Could you point me to the conversion code which converts the model from (https://github.com/google-research/vision_transformer) to vit_base_patch16_224.pth? It would be very helpful.

owaisCS avatar Mar 12 '22 10:03 owaisCS

Hi, @congee524 , can you find the scripts for converting the model from (https://github.com/google-research/vision_transformer)?

kennymckormick avatar Mar 20 '22 15:03 kennymckormick

Sorry, I cannot find the script. Maybe you can print the keys of the original model and the converted model, and then compare the difference. The key value should not change much (add and change prefix).

congee524 avatar Mar 22 '22 02:03 congee524

OK Thanks, @congee524, I will look into it. By any chance do you have the TimeSformer model pre-trained on ImageNet-1K??

owaisCS avatar Mar 23 '22 07:03 owaisCS

Hi @owaisCS , I have same question. Did you convert other models from https://github.com/google-research/vision_transformer to mmaction2's type? I tried to convert and read the keys from the original models, but it appears that the keys are distinct and have additional variables. Thank you.

daitranskku avatar Nov 14 '22 06:11 daitranskku