perception-traffic-light-detection

Introduction

After the previous projection in the preprocessing stage, a projection frame is obtained on the picture, but the obtained projection frame is not completely reliable, so a larger region of interest (Region of Interest ROI) to be calculated by the projected signal light position is used. to determine the exact bounding box of the semaphore. Signal light detection (detect) is a conventional convolutional neural network detection task, which receives images with ROI information as input data and sequentially outputs bounding boxes.

Directory Structure

├── traffic_light_detection // trafficlight detection module
    ├── conf            // module configuration files
    ├── dag             // dag files
    ├── data            // model params
    ├── detection       // main part for detect
    ├── interface       // function interface folder
    ├── launch          // launch files
    ├── proto           // proto files
    ├── traffic_light_detection_component.cc // component interface
    ├── traffic_light_detection_component.h
    ├── cyberfile.xml   // package management profile
    ├── README.md
    ├── README_cn.md
    └── BUILD

Modules

TrafficLightDetectComponent

apollo::perception::onboard::TrafficLightDetectComponent

Input

Name	Type	Description	Input channal
`frame`	`apollo::perception::onboard::TrafficDetectMessage`	trafficlight message	/perception/inner/Detection

Note: The input channel is structure type data. The default trigger channel is /perception/inner/Detection. The detailed input channel information is in modules/perception/traffic_light_detection/dag/traffic_light_detection.dag file. By default, the upstream components of the messages received by the component include traffic_light_region_proposal.

Output

Name	Type	Description	Output channal
`frame`	`apollo::perception::onboard::TrafficDetectMessage`	trafficlight message	/perception/inner/Retection

Note: The output channel is structure type data. The message is defined in the modules/perception/common/onboard/inner_component_messages/traffic_inner_component_messages.h file. The output channel message data can be subscribed by components in the same process. The detailed output channel information is in modules/perception/traffic_light_detection/conf/traffic_light_detection_config.pb.txt file.

configuration

component configuration files： modules/perception/traffic_light_detection/conf/traffic_light_detection_yolox_config.pb.txt

parameter type	parameter name	default value	meaning
string	PluginParam.name	TrafficLightTLDetectorYolox	detection algorithm name
string	PluginParam.config_path	perception/traffic_light_detection/data	configuration file path
string	PluginParam.config_file	detection_yolox.pb.txt	profile name
string	detection_output_channel_name	/perception/inner/Retection	detection result output channel
int32	gpu_id	0	gpu id

Yolox model configuration file location：modules/perception/traffic_light_detection/data/detection_yolox.pb.txt

parameter type	parameter name	default value	meaning
string	name	tl_detection_yolox	model name
string	framework	Onnx -> TensorRT	model inference framework
string	ModelFile.proto_file	yolox.onnx	model network structure
string	ModelBlob.inputs	images [1,3,384,384]	model input data name and dimension
int32	ModelBlob.outputs	bbox、conf、cls	model output data name and dimension
int32	max_batch_size	3	Multi-batch inference input model detection dimension
int32	min_crop_size	400	Crop box size
int32	classify_resize_width	384	model input image resize width
int32	classify_resize_height	384	model input image resize height

How to Launch

Add vehicle parameter configuration file to modules/perception/data/params in profile, corresponding frame_id and sensor_name, launch transform

bash

cyber_launch start /apollo/modules/transform/launch/static_transform.launch

Launch trafficlight perception

bash

cyber_launch start /apollo/modules/perception/launch/perception_trafficlight.launch