An automatic garbage detection using optimized YOLO model
Document Type
Article
Publication Date
2-1-2024
Abstract
Garbage pollution is an increasing global concern. Hence, the adoption of innovative solutions is important for controlling garbage pollution. In order to develop an efficient cleaner robot, it is very crucial to obtain visual information of floating garbage on the river. Deep learning has been actively applied over the past few years to tackle various problems. High-level, semantic, and advanced features can be learnt by deep learning models based on visual information. This is extremely important to detect and classify different types of floating garbage. This paper proposed an optimized You Only Look Once v4 Tiny model to detect floating garbage, mainly by improving the spatial pyramid pooling with average pooling, mish activation function, concatenated densely connected neural network, and hyperparameters optimization. The proposed model shows improved results of 74.89% mean average precision with a size of 16.4 MB, which can be concluded as the best trade-off among other models. The proposed model has promising results in terms of model size, detection time and memory space, which is feasible to be embedded in low-cost devices.
Keywords
Computer vision, Debris, Deep learning, Image processing, Object detection
Divisions
sch_ecs
Funders
Universiti Malaya [Grant No: IMG001-2022]
Publication Title
Signal Image and Video Processing
Volume
18
Issue
1
Publisher
Springer London Ltd
Publisher Location
236 GRAYS INN RD, 6TH FLOOR, LONDON WC1X 8HL, ENGLAND