search

fjchange / object_centric_VAD

An Tensorflow Re-Implement of CVPR 2019 "Object-centric Auto-Encoders and Dummy Anomalies for Abnormal Event Detection in Video"
MIT License
95 stars 30 forks source link

Only CAE perform better than CAE + Kmeans + SVM？ #11

Open Arbitersc opened 4 years ago

Arbitersc commented 4 years ago

hi, i run your code in my server and i found there is a 'test cae perform' function in the code. So, i run it, and got ''' 0.7018773380308058 dataset = shanghaitech, loss file = ../anomaly_scores/shanghaitech_CAE_only.pkl, auc = 0.6865759042248505 ''' And then i run the 'train_svm' function with the config below ''' bn: False class-add: False norm: False ''' and i got only the auc of all videos '0.67840248'. I think there are something wrong in my code when i trained svm. So i wanna know your result of the cae to prove it. Thank you.

lss0510 commented 3 years ago

Hello, is your result the same on the avenue data set?