Images with the same aspect ratio as 360×640. This model detect lane from inputs.
- input image
(Image from https://github.com/TuSimple/tusimple-benchmark)
Images which the detected lane in input images is colored in green.
- output image
Four green lines mean each detected lane which model predicted and numbers attached to each line mean line number.
Predicted curve parameters are showed at the top of the image.
You have to specify input filetype and input filepath.
- Image mode (image to image)
You run sample script as below if your desired file path is {Path to LSTR}/input/image/1.jpg.
$ python3 lstr.py --input input.jpg- Video mode (video to video)
You run sample script as below if your desired file path is {Path to LSTR}/input/video/video1.mp4.
$ python3 lstr.py -v input.mp4-
Repository
LSTR -
Input images and videos
Input images are part of TuSimple dataset and input videos are created by using TuSimple images.
PyTorch 1.8.0
ONNX opset = 11