Overview
Description
MosChip® solution for ADAS brake alert features distance monitoring and is tailored for deployment on RZ/V series vision MPUs with DRP-AI accelerators. This application offers a comprehensive solution for ADAS collision alert systems, focusing on detecting objects in the front view of a vehicle and monitoring the distance between the vehicle and detected objects. Inferenced YOLOv3 architecture + distance monitoring algorithm and is now deployed on the RZ/V2MA edge AI accelerator.
Image
Features
- Combination of pretrained YOLOv3 and TinyYolov3 and custom-trained algorithms for object detection
- Custom object detection models created using Pytorch framework, trained, validated and subsequently ported to ONNX format for compatibility and deployment purposes
- Pixel-based distance calculation algorithm developed and rigorously tested, demonstrating an accuracy rate of 60%
- For example, inference tasks on Edge AI accelerators achieve an impressive 23 ms inference time for TinyYolov3, delivering a corresponding speed of 10 frames per second (FPS) for detected classes
Comparison
Applications
- ADAS
- Distance measuring applications