This is my solution for the second in-class assignment, for the graduate level course COE-595 Advanced Computer Vision Processing, for the term T252, at KFUPM, taken as part of coursework for the (direct) PhD program in Computer Engineering.
We were expected to use YOLO26n, to train and run inference on a small PPE (Personal Protective Equipment) safety dataset and report the accuracy results. Additionally we were expected to alter the model, for which the techniques I used were based on the intuition for improving the model for the smaller classes that were showing weaker accuracy scores.
The assignment was in-class, with an additional window of 24 hours for code submission. Any kind of resources were allowed to be used without copy-pasting from any of them and it was obligated to write out the code in all cases. Groups of two were allowed for the purpose of team collaoboration and mutual help and learning.