A Distance-transform based arrow direction estimator
In traditional arrow direction estimators, it is difficult for single algorithms to estimate arrows with both 7 and 9 sides. This arrow direction estimator based on distance-transform methods provides a robust arrow direction estimator.
This is one of the research output of System Design Project 2022-2023 (Group 13), School of Informatics, The University of Edinburgh.
The algorithm performs the best when the arrow image is tightly bounded with the image frame (as we customise this algorithm with yoloV5 detected arrows with a bounding box).
OpenCV, numpy
python3 detect.py <image_filename>
Even worked in some extreme cases!
