The purpose of this project is to track staff members wearing a staff name tag.
Originally the plan was to use a variation of centroid tracking from openCV tracker algorithms such as 'KCF' or 'MOSSE', however it proved to be ineffective. So a deep learning model was used instead. The model that was used is YOLOv8m. To train the model, several reference images from sample.mp4 as well as several real-life pictures of myself. This was stored in /data. It was then annotated using labelImg. To see if the model was properly trained, a short video called sampleIRL.mp4 was tested.
The trained video of sample.mp4 can be found in the following directory: /runs/detect/predict2/sample.mp4 Whereas the trained video of sampleIRL.mp4 can be found: /runs/detect/predict/sample.mp4. When executing the program, please note that getting the frame at which the staff is present and the xy locations of staff members is a little slow due to the size of the array.
Example of the final output:
