Searching for Uncollected Litter with Computer Vision

TL;DR

This paper develops a computer vision approach to detect uncollected litter in images, aiming to assist waste management by mapping litter distribution, though challenges remain with different camera perspectives.

Contribution

It introduces a method combining photo metadata and computer vision to identify litter categories, highlighting the need for diverse datasets to improve performance across perspectives.

Findings

Effective detection with smartphone photos

Reduced accuracy with vehicle-mounted camera images

Diverse datasets are necessary for better generalization

Abstract

This study combines photo metadata and computer vision to quantify where uncollected litter is present. Images from the Trash Annotations in Context (TACO) dataset were used to teach an algorithm to detect 10 categories of garbage. Although it worked well with smartphone photos, it struggled when trying to process images from vehicle mounted cameras. However, increasing the variety of perspectives and backgrounds in the dataset will help it improve in unfamiliar situations. These data are plotted onto a map which, as accuracy improves, could be used for measuring waste management strategies and quantifying trends.