A SUAI student has proposed a new approach for managing household waste in a megalopolis. It is based on a neural network pipeline that is able to determine the trash can in the image and its level of fullness
The neural network pipeline consists of two neural network models: a garbage can detector and a classifier that determines the level of filling of a garbage tank. The author of the study collected images of dumpsters of varying degrees of occupancy, and then selected a neural network model for detecting a dumpster in the frame and retrained the selected model.
This solution will make it possible to automate the application moderation process and optimize the garbage collection process. If there are city surveillance cameras in the field of view of which container sites with garbage cans fall, this data can also be used to detect both full and empty (half-empty) garbage cans. This information will optimize the routes of sanitation truckы by eliminating those containers that do not require emptying, and adding those that are already filled, but were not included in the original garbage collection schedule.
