Download the YOLOv3 weights here. Use the YOLOv3-416 as that has a good mAP with over 30 fps. Put it in Yolo/weights. The corresponding cfg file is already included.

Download the LaneNet weights here. Download all the files in the folder and put it in LaneNet/weights.

LaneNet parameters can be changed in the LaneNet/config/global_config.py file. The last three postprocessing parameters are the most relevant ones to find the optimal accuracy/speed tradeoff.

Put images you want to test in the Images/input folder. Then go to the command line, cd to this directory and run python main.py --input images/input --output images/output --yolo_weights Yolo/weights --lanenet_weights LaneNet/weights/tusimple_lanenet_vgg.ckpt.

The object detection output are be displayed in Images/Output/objects, the lanes are displayed in Images/Output/lanes, and the JSON outputs are in Images/JSON.