dswelch
commented
2 years ago Working on building OpenCV with cuda so that we can possibly use the GPU to speed things up. Once we have that, most opencv methods are available with cuda so we shouldn't have to change much, but just need to fix up whatever files we want to use the GPU for (probably ADSDetection.py only). Also looked into using dst_points in the function that converts to occupancy grid, rather than having to draw in the unknown areas and then converting.
Took hertz measurements and put them in the "Run Commands" file in the drive. Tried to speed up the perspective transform node by not doing the math every time but it didn't really speed it up. Next step is to see if there is anything to optimize in the lane detection algorithm.