search

WaqasSultani / AnomalyDetectionCVPR2018

503 stars 183 forks source link

loss when training #28

Open LakyTT opened 5 years ago

LakyTT commented 5 years ago

Hi,WaqasSultani What is the loss value when you train the network?I don't know whether my loss value is right.It's 0.29 after training。 Thanks

LakyTT commented 5 years ago

And your code Test_Anomaly_Detector_public.py seems have no module to save result(*.mat) of test.

LakyTT commented 5 years ago

We know，abnormal segments have high scores,normal segments low scores.But when you test,how do you define the scores?In other words,you think it is abnormal,when score is more than threshold.what's the threshold?

WaqasSultani commented 5 years ago

Hi, On testing, you evaluate at multiple threshold and plot ROC curve and compute AUC

LakyTT commented 5 years ago

Thanks for your reply. It seems you draw figure with matplotlib when jj=i%25(in code Demo.py).That's to say you draw a figure for every 25 frames?Could you tell me the details of assigning the score to each frame? And I find the score may be greater than 1 and less than 0？

Thanks