How to get the test accuracy?

[wyl0608]

First of all, thanks for sharing, and I have some questions:

1. In test.py, I make threshold=0.00055,some normal events and some abnormal events were obtained, but the total number of events was only 24. The code in processor.py is to extract 1 frame every 5 seconds and I change to extract 5 frames every 1 second, so that about 2 seconds as a bunch (event) ,is it more reasonable ?
2. How do I get the test accuracy in this work, and how do you do it? Do I need to write a module for the reconstruction error of each frame to get the regularity score of the entire data set?
3. In processor.py, at line 76 "for video in videos[1:]:" seems to have lost a video. Looking forward to your reply

[harshtikuu]

@wyl0608 haven't written a module for test accuracy. Will add it soon. Also videos[1:] was to deal with an extra file that was somehow created in my OS. You can alter it!

[wyl0608]

thanks for the reply,waiting for the module.

[arash9900]

I am also waiting for the test accuracy module,will it happen soon?

Thanks

[PBRAOS]

First of all, thanks for sharing, and I have some questions:

1. In test.py, I make threshold=0.00055,some normal events and some abnormal events were obtained, but the total number of events was only 24. The code in processor.py is to extract 1 frame every 5 seconds and I change to extract 5 frames every 1 second, so that about 2 seconds as a bunch (event) ,is it more reasonable ?

2. How do I get the test accuracy in this work, and how do you do it? Do I need to write a module for the reconstruction error of each frame to get the regularity score of the entire data set?

3. In processor.py, at line 76 "for video in videos[1:]:" seems to have lost a video.
   Looking forward to your reply

Good point on the 1/5 vs 5 !. I am about to check if it makes a difference on the calibrations.

[AndyHon]

thanks for the reply,waiting for the module.

How did you solve this problem?Can you share your code?