TrafficLight Classification module with YoloV7 detector

This source is only includes inference code. Our detector has a seven different classes:

• Person, Bike, Car, Bus, Truck, Traffic Sign, Traffic Light, Emergency Car Our traffic light classification supports a 12 traffic light types.
• green, green_left, off, other, pedestrain, red, red_left, red_yellow, yellow, yellow21, yellow_green4, yellow_other (South Korea).

base code of YoloV7

A PyTorch implementation of YOLOv7.

• Paper Yolov7: https://arxiv.org/abs/2207.02696
• Source code: https://github.com/WongKinYiu/yolov7

TrafficLight Classification trainer code

Our traffic light classification code (Link: https://github.com/AUTOSELFDRIVING2022/TrafficLight_Classification_Trainer)

0. Weight Download

• YoloV7 weight Drive(Link: https://drive.google.com/file/d/1_bop55xJF-W0p3l7S2RS6NTLUxqcfU3V/view?usp=sharing)
• Resnet34 Traffic Light Classification weight Drive(Link: https://drive.google.com/file/d/1EA_MMeS59hREHRm72qTVdeMOPVqjgfGI/view?usp=sharing)
• Save pretrained weight to this path: ./data/weight/

1. Inference

• Load pretrained yolov7 modle weights.
• Inference Traffic Light Classification:

python inference_YOLOv4_TL_Classification_TRACK.py --det_weight {weight_file_of_detector} --tl_weights {tl_classification_weight} --ec_weights {ec_classification_weight} --source {input_file_directory} --save_result_image {save_result_true_false}