search

zhengshou / scnn

Segment-CNN: A Framework for Temporal Action Localization in Untrimmed Videos via Multi-stage CNNs
Other
235 stars 102 forks source link

How to determine the train labels in untrimmed videos of proposal network? #17

Open buaa-luzhi opened 7 years ago

buaa-luzhi commented 7 years ago

@zhengshou Thank you very much! I have read it carefully! For peoposal networks, I was in trouble during the production of the training set. I don't know how to determine the candidate segment label. Is 1 or 0 for candidate segments label? I just want to know how to determine training set labels for candidate segments in untrimmed videos. Do you have any script implementation?

Thank you in advance!

Jillian2017 commented 7 years ago

@buaa-luzhi I cannot load the pretrained model from Dropbox. Would you send them to me, my email is zhoujing9204@163.com, thanks very much.

buaa-luzhi commented 7 years ago

@Jillian2017 it is too big