Skip to content
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

Added Haar cascade for russian cars licence plate detection #2777

Merged
merged 1 commit into from
May 29, 2014

Conversation

alexstadnik
Copy link
Contributor

Cascade works pretty fast, stump based, 16 stages overall.

@adrians
Copy link
Contributor

adrians commented May 21, 2014

I'm curious how does this compare to #2752 in terms of accuracy and performance.

Also, as a pro-tip: don't submit modifications inside your local "master" branch, because it'll go out of sync pretty fast. Try to work in another branch and submit a pull-request with that. This way, if you want to work on another patch you can branch again from your local master (sync'd with the upstream one), and not induce patch-dependencies in the new pull-request.

@alexstadnik
Copy link
Contributor Author

Hi!
In terms of performance, this one has less stages, so it definitely runs a bit faster. According to standart example:
results for cascade from #2752:
Processing 1 --cascade=haarcascade_russian_plate_number.xml
Processing 2 vlcsnap-2013-08-12-15h53m18s33.png
detection time = 298.194 ms
//
Processing 1 --cascade=haarcascade_russian_plate_number.xml
Processing 3 vlcsnap-2013-08-12-15h54m12s17.png
detection time = 279.121 ms
/////////
results for given cascade:
Processing 1 --cascade=haarcascade_pltzzz64x16_16STG.xml
Processing 2 vlcsnap-2013-08-12-15h53m18s33.png
detection time = 172.356 ms
//
Processing 1 --cascade=haarcascade_pltzzz64x16_16STG.xml
Processing 3 vlcsnap-2013-08-12-15h54m12s17.png
detection time = 174.658 ms
//////
As for accuracy, this matter a bit more complex: #2752
#2752 detects more plates, but it gives more false detections also. For scientific comparison ROC-curve should be presented, but its a lot of work to do.

My commit was initiated by #2752, as an alternative, the same way as now we have few cascades in library for face detection.

I'm attaching results of both cascades for 2 images, to clearify a bit accuracy of detection.

Thanks for the pro-tip, I've made a branch, should I recommit cascade from new branch if you approve?

Sincerely.

detection results for given cascade:
example_result1-0
example_result1-1

detection results for cascade from #2752:
example_result0-0
example_result0-1

P.S.
Cascade from #2752 doesn't load by cvLoadHaarClassifierCascade in v.2.4.6, but I'm not sure is this a real thing.

@vpisarev vpisarev self-assigned this May 27, 2014
@vpisarev
Copy link
Contributor

great, thank you! 👍 (hopefully, I won't start receiving more speeding tickets sometimes soon :) )

@opencv-pushbot opencv-pushbot merged commit 73bfcf8 into opencv:master May 29, 2014
SpecLad pushed a commit that referenced this pull request May 29, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

5 participants