search

POSTECH-IMLAB / Traffic_sign_ADAS

GNU Lesser General Public License v3.0
28 stars 30 forks source link

How to work this algorithm? #2

Closed bnkang closed 4 years ago

bnkang commented 6 years ago

I want to know this detection and recognition method in detail. Is there any reference papers or technical reports? If yes, where can I download these materials?

nyamnym commented 6 years ago

http://imlab.postech.ac.kr/research_web_ADAS_TrafficSign.htm

I hope this link can answer your question. This link explain about my traffic sign detection & recognition algorithm in detail.

bnkang commented 6 years ago

@nyamnym I have seen a link page you mentioned. I have two questions. 1) As you used ACF for feature extraction, what types and how many of features did you use for aggregation? how about feature dimension to train svm? 2) To detect traffic signs, you used boosting classifier. what boosting algorithms did you used for classifier? AdaBoost??

nyamnym commented 6 years ago

@bnkang

  1. ACF uses total 10 channels for aggregation. LUV channel ( 3 ) + unoriented gradient magnitude channel ( 1 ) +gradient channel for 6 direction ( 6 ) To train SVM, I used HOG feature with 8x8 block size, 4x4 cell size and 9 bin size. So I got 8100 feature dimension.
  2. I used Boosted Decision Tree algorithm for detecting.