search

hobincar / RecNet

A Pytorch implementation of "Reconstruction Network for Video Captioning", CVPR 2018
MIT License
54 stars 19 forks source link

How to extract 3D-ResNext Features? #19

Closed Daic115 closed 4 years ago

Daic115 commented 4 years ago

Hi hobincar,thanks for your excellent work！ I wonder how you extract the 3D-ResNext feature in your video-feature-extractor. I have tried using pretrained 3d model weight from here,and using feature-extractor under default setting, but at captioning stage ,the features which I extracted works far worse than your 3d-resnext features. Could you please tell me some details about your features extractor setting?

hobincar commented 4 years ago

Hi,

Maybe this repo can help you. Based on the original repo, I uploaded a commit containing the changes for extracting 3D-ResNext features I used.

Daic115 commented 4 years ago

Thank you very much!