kaist-avelab/K-Lane — reverse-engineered prompt

Please get this K Lane project into a usable state for me. I want a Python workspace where I can use the KAIST LiDAR lane dataset, run a pretrained lane detector on test data, train on the full dataset, and validate the results without having to guess the setup.

Please make the visualization tool work so I can inspect inference results and do LiDAR to camera calibration, and make sure the annotation tool is usable for labeling point clouds. Set up the expected folders, sensible default configs, and simple run commands so the whole flow works from dataset download through training and testing. If pretrained model files are available, wire those in too.

Assume I am on Ubuntu with a GPU. If anything in the repo is outdated or broken, fix it and use the current docs or look up missing details online if needed. Leave me with a short README that explains install, dataset setup, training, validation, and how to launch the GUI tools.