A study on the use of deep learning for trash classification
Published: July 30, 2026
Abstract
This paper presents an experiment on the use of a convolutional neural network (CNN) model to classify garbage into three different groups, namely plastic bottles, cans, and glass by using 900 manually collected images. In which, 500 images were used for analysis and 400 images were used for evaluation of results. This model gave a linear result which is between the amount of training data and the average accuracy (mean Average Precision, mAP). Based on these results, an other experiment was done to ensure that the trained model meets set requirements including AI401 with 80% overall accuracy.
Keywords
garbageconvolutional neural networkdeep learning modelopen sourceTensorflow Object Detection API
References
1.
VMENR (Vietnam Ministry of Environment and Natural Resources). (2014). Canadian International Development Agency, Vietnam Environment Monitor: Solid Waste, Vietnam.
2.
DONRE (Department of Natural Resource and Management)-Ho Chi Minh City. (2014). The report on overview of the SWM system of HCMC, Vietnam.
3.
Jiahui Yu, Yuning Jiang, Zhangyang Wang, Zhimin Cao, and Thomas Huang. Unitbox,. (2016). An advanced object detection network, In Proceedings of the 24th ACM international conference on Multimedia, ACM, pp. 516-520.
4.
Shi, Lin; Zhong, Hao; Xie, Tao; Li, Mingshu. (2011). An Empirical Study on Evolution of API Documentation, International Conference on Fundamental Approaches to Software Engineering, Lecture Notes in Computer Science, ISBN 978-3-642-19810-6. Retrieved 22 July.
