2024-07-15 We open the dataset and updated the documentation for training and deploying models.
2023-10-06 We released code and pre-trained model.
1、This project is an on-board assistance system based on target detection algorithm, which also has the ability to detect pedestrians and road defects.
2、The deployed hardware device is Raspberry Pi 4B.
3、The specific implementation effect is bilibili video
1、Model Training
The models used are yolov5 and yolov5-lite. For the specific training process, refer to YOLOv5-Lite: README.md
2、Model deployment
The framework used for deployment is NCNN. For specific steps, refer to test320_yolov5_lite: README.md
3、Visualization
Read the coordinates through GPS and call the Amap API to mark them. For details, refer to Visualization
4、Distance measurement module
The ranging module is divided into ultrasonic ranging and visual ranging. For details, please refer to Distance
[1] Xiangrong Chen, & Ziman Gong. YOLOv5-Lite: Lighter, faster and easier to deploy.