POSTECH-IMLAB / traffic_light_recognition

traffic light recognition system for ADAS

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Traffic Light Recognition (TLR)

This TLR system designed as a module of Advanced Driver Assistance Systems (ADAS). It detects and recognizes traffic lights from driving-view image sequences.

Hyundai Contest (2017)

TLR system designed for Hyundai Contest (2017). TLs are detected by using ACF and HUV-Histogram.

Accurate Traffic Light Detection using Deep Neural Network with Focal Regression Loss (2018)

Based on RGB camera, implmented with Caffe. For training from scratch:: train -solver tlr_yolov2_deconv_resnet101_focalloss4_4.prototxt

For training from weights:: train -solver tlr_yolov2_deconv_resnet101_focalloss4_4.prototxt -weights deconv_focal_tlr_iter_62000(best).caffemodel -gpu 0

About

traffic light recognition system for ADAS

License:GNU Lesser General Public License v3.0


Languages

Language:Python 79.4%Language:C++ 18.0%Language:Shell 2.3%Language:Dockerfile 0.3%