Details in feature extractor (in temporal dimension)

Roc-Ng / DeepMIL

Real-world Anomaly Detection in Surveillance Videos CVPR2018 UCF-Crime dataset

119 stars 11 forks source link

Details in feature extractor (in temporal dimension) #9

Open sdjsngs opened 3 years ago

sdjsngs commented 3 years ago

Hi i have some problem in feature extractor in this paper. assume one video , its temporal length is T so you split it to 32 segments frtist as the paper said? segment_length is T//32 and do feature extactor in slide windows in step 1 ? so each segment would get T//32-15 features and merge them to one feature ?
Looking forward to your reply

Roc-Ng commented 3 years ago

please see the function "process_feat" in utils.py. Thx.