search

ktr-hubrt / WSAL

Official codes for paper: Localizing Anomalies from Weakly-Labeled Videos
73 stars 9 forks source link

How to extract the 32*1024D feature? #14

Open zimengxueying opened 2 years ago

zimengxueying commented 2 years ago

"we use the model weights finetuned on UCF-Crime as in [5] to extract features. While on our TAD dataset we only use the model weights pretrained on Kinetics-400 dataset." The link to GCN https://github.com/jx-zhong-for-academic-purpose/GCN-Anomaly-Detection/tree/master/feature_extraction make me confused about it because it provided ucf-crimes -1000 and 1400 for rbg and flow ,how could I used it? If you could help me ,I'll feel so grateful!

borijang commented 2 years ago

Hi, I am facing the same problem. Did you find out how to make this work?

zimengxueying commented 2 years ago

Hi, I am facing the same problem. Did you find out how to make this work?

Not yet.. What about you? Do you also do research about Weakly supervised anomaly detection ？ Would you willing to do more communication? Thanks so much!