annusha/lirec

Learning Interactions and Relationships between Movie Characters (CVPR'20)

lirec's Issues

hi,dataset query ! (●'◡'●)

hi , beautiful author ~:

Your work is perfect and interesting . How can we get the bbox of a whole person (raw dataset is only give person of face) ?

Hope to get your reply ~

Best wishes

jun0

Problem on Feature Extraction

Mentioned in paper, visual feature extraction part:

Visual features. We extract visual features for all clips using a ResNeXt-101 model [20] pre-trained on the Kinetics-400 dataset. A batch of 16 consecutive frames is encoded, and feature maps are global average-pooled for the clip represen- tation, and average pooled over a region of interest (ROIPool) to represent characters. Given a trimmed clip v i, we max pool above extracted features over the temporal span of the clip to pick the most important segments.

And I followed with this tool to extract visual features, but for any video data in MovieGraph dataset, feature value still mismatch!
Could you tell more processing details?

Please kindly reply for my doubts.
Thanks!

hi,dear author about dataset

Dear author，
I'm sorry to bother you again. When I used the original data set to extract more information features, I found that the faces of some videos could not match the faces of the videos themselves. How did you deal with them?
Hope you can reply me .. I want to graduate /(ㄒoㄒ)/~~

Best wishes，
jun0

raw video data

Dear author:
Hello, I have downloaded the MovieGraphs dataset, but there is no original video in it. How did you get the original video? We look forward to your reply.
Thank you very much.

issue with finding modules

I have done all the steps up until step 3 but when I run "python resume/modalities.py", I got an error indicating "No module named 'mixed_utils'. Could you please help me what can be the problem?