This method use a fine-tuned CLRNet model as the detector, but users can change it to the detector they use.

• Python >= 3.8 (tested with Python3.8)
• PyTorch >= 1.6 (tested with Pytorch1.6)
• CUDA (tested with cuda10.2)
• Other dependencies described in requirements.txt

To install required dependencies run:

$ pip install -r requirements.txt

To run the tracker with the provided detections:

$ cd path/to/project
$ python lanetracker.py

And the results of assigning instance IDs are saved in output

To show the visible results you need to:

1. Use your own private dataset. Or contact us and we will send you the download address of our own dataset.

2. Create a symbolic link to the dataset

$ ln -s /path/to/dataset mot_benchmark

3. Run the demo with the --display flag

$ python lanetracker.py --display

To calculate lane offset and show the result:

$ cd demo
$ python calc_offset.py

The trajectory of the vehicle will be saved in demo/lane_offset_plot.png. It's corresponding visible result on our private dataset compared with the counterpart of the baseline method shows in demo/demo.avi.