Project Lead: Wenting Duan; Project Student RAs: Oakleigh Weekes, Kyle Fogarty.
The allocation of resources to combat litter is currently a manual process. However, dash-cameras are being more common in motoring, and potentially offer a mechanism by which litter can be more thoroughly and actively monitored.
The Green Verge project seeks to implement a robust system that can automatically detect litter from dashcam footage, register the detected litter and geospatial coordinates, and produce a user-friendly mapping solution that highlights regions of high litter concentration.
A YoloV5s model was trained on over 16k augmented images generated from over 8k dashcam frames. A baseline model was achieved which can detect litter, plotting routes on folium maps was possible through obtaining location data via optical character recognition of dashcam footage, with areas of dense litter being linked to locations. For more details, please see the following links.
To find out more about the project, check out the GitHub page: https://github.com/Green-Verges-University-of-Lincoln