ARV HSV Tuner

by Matthew Gawthrop

Algorithm description

Utilizes hsv filtering and YOLO ML models to gain a consistent occupancy grid of mainly lane line, potholes, and traffic barrels.

hsv.py is designed to be used as a library, meaning the file is meant to be copied into other projects and then imported into a file for use from hsv import hsv. Each object of the hsv class is tied to a video path. This video path is not only used for saving of hsv filtering values but also is directly used for streaming video during tuning.

Tuning

Reference demo.py for a look at the functionality with tuning.

You must first create an object with the video path that you want to tune for (examples/hsv_data_template.json). Then by calling hsv_obj.tune(mask_name) you can slide the sliders back and forth until you reach the mask you want. Finally, slide the "Done Tuning" slider and the values will be saved in <your_defined_hsv_name>.json. You can do this over and over with new mask_names to get a whole suite of masks that are connected to one video path. Change the video path on construction of the object and you can create a new set of masks for that video too.

Once you have completed all your tuning, use get_mask(frame) to get the individual and combined masks you tuned for.

Getting Occ Grids

Use combined, mask_dictionary = get_mask(frame) which will return a 2D numpy array of all combined masks that have been tuned for this object. It also returns a dictionary which maps from mask_name -> occ grid which can be used to get just one colors occupancy grid at a time.