This package contains a bunch of computer vision nodes that use OpenCV and rospy. Each node is a subclass of a superclass called LaneDetection, which initializes publishers and subscribers, as well as dynamic_reconfigure callbacks, and converts ROS to CV2 and vice versa. All the nodes subscribe to a CompressedImage topic and publish to a CompressedImage topic. Some nodes require either grayscale or RGB images (backproject needs RGB. fitline, brightest-pixel, and skeletonize need grayscale).
You must use launch files to run nodes (particularly because of the roi.yaml parameters).
Just use generic.py as a template. Replace the word "generic" with whatever name your node is (watch the underscores or uppercase letters!). Add some code to the middle of image_callback to make your node do stuff (roi is your input image, and final_image is your output image). Use the convert_to_mono function if necessary.
Multiple nodes can be chained in launch files (their primary purpose is to test different combinations of filters to help find the best algorithm for detecting white lines on grass).
Check the generic.launch and stereo-generic.launch files for the default parameters.
If you've used and built very old versions (before LaneDetection class) of line_detection then tried to use the newer versions, you might have trouble with the dynamic_reconfigure node crashing because of some cfg not found error. Unfortunately, it seems this problem persists until you remove your entire workspace and start fresh (perhaps even have to reinstall ROS completely [?]). It has to do with some weird cfg filename changes persisting even after attempting clean builds. Just stay away from the old versions.
pixel\_to\_coordinate\_calculatornode:image_geometrymight have to be installed and compiled from source. Check this issue for more details.