As part of my undergraduate university robotics's team -- RUDRA, I created a DL model that can detect pictures of arrows and in which direction they are pointing to. This repository contains all the files related to my project on detecting arrows using pre-trained neural networks. I used TensorFlow 2's Object Detection Module primarily but also tried and tested Yolo with the help of my teammate (had better inference speeds but can't remember why we didn't use it). Interesting fact: this was pains taking work for me. I spent way more time on developing it than I should have as I didn't have any prior knowledge or expertise when I started this work. Literally created my own dataset by holding a printout of an arrow. Not once. Not twice. But thrice. And even annotated 1000+ images each time. Altough the detection didn't work as expected by the day of the competition, still got to learn a ton by doing this project.
Also check out this stop motion video of my dataset: https://youtu.be/WgvSXuW0MkQ
The weights are not included in the repo because the file size is too large.