-
Notifications
You must be signed in to change notification settings - Fork 431
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[detectors] Add Koala-36M Based Detector #459
base: main
Are you sure you want to change the base?
Conversation
b66d897
to
6b0aef9
Compare
c691469
to
82d160a
Compare
82d160a
to
fe1de64
Compare
fe1de64
to
700b27d
Compare
700b27d
to
cb9c71e
Compare
Implement algorithm similar to that described in Koala-36M. Add `KoalaDetector` and `detect-koala` command. #441
cb9c71e
to
1b757a3
Compare
@@ -37,6 +37,7 @@ The following results indicate that ContentDetector achieves the highest perform | |||
| HashDetector | 92.96 | 76.27 | 83.79 | 16.26 | | |||
| HistogramDetector | 90.55 | 72.76 | 80.68 | 16.13 | | |||
| ThresholdDetector | 0.00 | 0.00 | 0.00 | 18.95 | | |||
| KoalaDetector | 86.83 | 78.38 | 82.39 | 97.75 | |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, interesting. Even though Koala detector uses ML-based methods, it does not achieve good performance compared with rule-based detectors?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's possible I have a bug somewhere in the implementation. This was a first-pass at implementing this with just a single video test case. I haven't reviewed the code in some time - if you spot anything that looks wrong, please let me know!
Unfortunately I haven't had time to work on this PR lately. If you want to develop this detector further, I would be happy to clean this PR up and get it ready for review.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I see. To be honest, I am not sure about the performance of Koala-37M because the authors do not evaluate the shot detection accuracy in the paper. Because they released the code (https://github.com/KwaiVGI/Koala-36M/blob/main/trainsition_detect/VideoTransitionAnalyzer.py), let me check the performance on BBC and AutoShot. I will report it in this thread.
Work in progress. #441
Based off of https://koala36m.github.io/ (will update this with proper citation when ready for review)