Benchmarking computer vision models for automated construction waste sorting

IF 11.2 1区环境科学与生态学 Q1 ENGINEERING, ENVIRONMENTAL Resources Conservation and Recycling Pub Date : 2024-11-23 DOI:10.1016/j.resconrec.2024.108026

Zhiming Dong, Liang Yuan, Bing Yang, Fan Xue, Weisheng Lu

{"title":"Benchmarking computer vision models for automated construction waste sorting","authors":"Zhiming Dong, Liang Yuan, Bing Yang, Fan Xue, Weisheng Lu","doi":"10.1016/j.resconrec.2024.108026","DOIUrl":null,"url":null,"abstract":"<div><div>Waste sorting is a critical process in construction waste management system. Computer vision (CV) offers waste sorting automation potential by recognizing waste composition and instructing robots or other mechanical devices accordingly. However, how the plethora of CV models developed perform relative to each other remains underexplored, making model selection challenging for researchers and practitioners. This study aims to benchmark existing CV models towards automated construction waste segregation. Seventeen models were selected and trained with unified configuration, and then their performance was evaluated on the aspect of accuracy, efficiency, and robustness, respectively. In experimental results, BEiT attained top accuracy (58.31 % MIoU) while FastFCN had the best efficiency (12.87 ms). SAN displayed the least standard deviation (4.41 %) for robustness evaluation. This research contributes a reliable reference for CV model selection, advancing automated construction waste sorting research and practices, and ultimately promoting efficient recycling while reducing the environmental impact of construction and demolition waste.</div></div>","PeriodicalId":21153,"journal":{"name":"Resources Conservation and Recycling","volume":"213 ","pages":"Article 108026"},"PeriodicalIF":11.2000,"publicationDate":"2024-11-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Resources Conservation and Recycling","FirstCategoryId":"93","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0921344924006177","RegionNum":1,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, ENVIRONMENTAL","Score":null,"Total":0}

引用次数: 0

Abstract

Waste sorting is a critical process in construction waste management system. Computer vision (CV) offers waste sorting automation potential by recognizing waste composition and instructing robots or other mechanical devices accordingly. However, how the plethora of CV models developed perform relative to each other remains underexplored, making model selection challenging for researchers and practitioners. This study aims to benchmark existing CV models towards automated construction waste segregation. Seventeen models were selected and trained with unified configuration, and then their performance was evaluated on the aspect of accuracy, efficiency, and robustness, respectively. In experimental results, BEiT attained top accuracy (58.31 % MIoU) while FastFCN had the best efficiency (12.87 ms). SAN displayed the least standard deviation (4.41 %) for robustness evaluation. This research contributes a reliable reference for CV model selection, advancing automated construction waste sorting research and practices, and ultimately promoting efficient recycling while reducing the environmental impact of construction and demolition waste.

Abstract Image

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

建筑垃圾自动分类计算机视觉模型基准测试

垃圾分类是建筑垃圾管理系统中的一个关键过程。计算机视觉（CV）通过识别废物成分并相应地指示机器人或其他机械设备，为废物分类自动化提供了潜力。然而，对已开发的大量 CV 模型的性能如何进行比较的探索仍然不足，这使得研究人员和从业人员在选择模型时面临挑战。本研究旨在对现有的 CV 模型进行基准测试，以实现建筑垃圾的自动分类。研究选取了 17 个模型，并对其进行了统一配置训练，然后分别从准确性、效率和鲁棒性三个方面对其性能进行了评估。实验结果表明，BEiT 的准确率最高（MIoU 为 58.31%），FastFCN 的效率最高（12.87 毫秒）。在鲁棒性评估中，SAN 的标准偏差最小（4.41 %）。这项研究为 CV 模型选择提供了可靠的参考，推动了建筑垃圾自动分拣研究和实践，最终促进了有效的回收利用，同时减少了建筑和拆除垃圾对环境的影响。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文去求助

来源期刊

Resources Conservation and Recycling 环境科学-工程：环境

CiteScore

22.90

自引率

6.10%

发文量

625

审稿时长

23 days

期刊介绍： The journal Resources, Conservation & Recycling welcomes contributions from research, which consider sustainable management and conservation of resources. The journal prioritizes understanding the transformation processes crucial for transitioning toward more sustainable production and consumption systems. It highlights technological, economic, institutional, and policy aspects related to specific resource management practices such as conservation, recycling, and resource substitution, as well as broader strategies like improving resource productivity and restructuring production and consumption patterns. Contributions may address regional, national, or international scales and can range from individual resources or technologies to entire sectors or systems. Authors are encouraged to explore scientific and methodological issues alongside practical, environmental, and economic implications. However, manuscripts focusing solely on laboratory experiments without discussing their broader implications will not be considered for publication in the journal.