Example scripts for the detection of lanes using the ultra fast lane detection model in ONNX.

Source: https://www.flickr.com/photos/32413914@N00/1475776461/

For performing the inference in Pytorch, check my other repository Ultrafast Lane Detection Inference Pytorch.

• Jetson Xavier AGX. JetPack 4.6

• OpenCV, scipy, onnx and onnxruntime. pafy and youtube-dl are required for youtube video inference.

• OpenCV, scipy, onnx and onnxruntime. pafy and youtube-dl, Nvidia Xavier AGX, JetPack 4.6 and Python3.6 .

pip install -r requirements.txt
pip install pafy youtube-dl

pip3 install opencv-python
pip3 install scipy
Download >> https://nvidia.app.box.com/s/bfs688apyvor4eo8sf3y1oqtnarwafww
Install  >> pip3 install onnxruntime_gpu-1.8.0-cp36-cp36m-linux_aarch64.whl
pip3 install scikit-build
if you found not match version try to upgrade the PIP >> sudo -H pip3 install --upgrade pip

The original model was converted to different formats (including .onnx) by PINTO0309, download the models from his repository and save it into the models folder.

ONNX Conversion script: cfzd/Ultra-Fast-Lane-Detection#218

The pretrained Pytorch model was taken from the original repository.

• Input: RGB image of size 800 x 200 pixels.
• Output: Keypoints for a maximum of 4 lanes (left-most lane, left lane, right lane, and right-most lane).

• Image inference:

python imageLaneDetection.py 

• Webcam inference:

python webcamLaneDetection.py

• Video inference:

python videoLaneDetection.py

• YoutTube: https://www.youtube.com/watch?v=yBmcYDke7Wg
• Original video: https://youtu.be/2CIxM7x-Clc (by Yunfei Guo)