The paper caculates error based on regularity score

[santhoshkelathodi]

When I tested on avenue dataset, I found that the code does not detect anomaly of

1. a person running fast on the walk way (avenue_01 test video).
2. The loss value was higher when a child appears on the left bottom corner or when a women starts walking towards the walkway appearing suddenly on the scene.

is it because anomaly is not based on regularity score?

[Anothorld]

Hi, could you tell me how do you preprocess the data? I use the code to preprocess the data on avenue dataset and train the model but found all the test video was detected as normal.