ISSN: 0304-128X ISSN: 2233-9558
Copyright © 2025 KICHE. All rights reserved

Articles & Issues

Conflict of Interest
In relation to this article, we declare that there is no conflict of interest.
articles This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/bync/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright © KIChE. All rights reserved.

Articles in press

폐플라스틱 재활용을 위한 최신 객체 탐지 모델의 성능 분석 : Swin transformer와 비전 모델 평가

Performance Analysis of State-of-the-Art Object Detection Models for Plastic Waste Recycling : Evaluation of Swin Transformer and Vision Models

Kim Doeun1 Junhyeok Son1 Dogyeong Kim1 Yuchan Ahn1†
1계명대학교 화학공학과
In Press, Journal Pre-proof, Available online 1 February 2026

Abstract

본 연구는 폐플라스틱 자동 분류를 위한 최신 인공지능 객체 탐지 모델의 성능을 평가하였다. Faster R-CNN, YOLOv8, YOLOv11, Swin Transformer 총 4개 모델을 대상으로 PET, PS, PP, PE 4종의 폐플라스틱 이미지 48,000개로 학습 및 테스트를 수행하였다. Optuna를 활용해 하이퍼파라미터를 최적화하고, 정확도, 정밀도, 재현율, F1 score, 평균 정밀도 평균값 (mAP), 추론 시간 등으로 성능을 평가하였다. Swin Transformer는 최고 정확도 (0.988)와 mAP (0.988)를 기록하며 복잡하거나 오염된 플라스틱에서도 우수한 분류 성능을 나타냈다. YOLOv11은 가장 빠른 추론 속도 (61.67 ms/이미지)를 보이며 실시간 처리 환경에 적합함을 확인하였다. 연구 결과, 모델 선택 시 정확도와 처리 속도 간의 트레이드오프가 존재함을 확인하였으며, 폐플라스틱 재활용을 위한 AI 기반 분류 시스템 구축에 실질적인 참고 자료를 제공한다.
This study evaluates the performance of state-of-the-art AI-based object detection models for automated plastic waste classification. Four models—Faster R-CNN, YOLOv8, YOLOv11, and Swin Transformer—were trained and tested on a dataset of 48,000 images covering four plastic types (PET, PS, PP, PE). Hyperparameters were optimized using Optuna, and performance was assessed via accuracy, precision, recall, F1 score, mean Average Precision (mAP), and inference time. Swin Transformer achieved the highest accuracy (0.988) and mAP (0.988), demonstrating superior classification capability, particularly for complex or contaminated plastics. YOLOv11 exhibited the fastest inference speed (61.67 ms/image) with competitive accuracy, highlighting its suitability for real-time applications. These results reveal a trade-off between accuracy and processing speed, providing guidance for model selection based on application requirements. This study offers quantitative benchmarks and practical insights for deploying AI-driven plastic recycling systems.

The Korean Institute of Chemical Engineers. F5,119, Anam-ro, Seongbuk-gu, Seoul, Republic of Korea
Phone No. +82-2-458-3078FAX No. +82-507-804-0669E-mail : kiche@kiche.or.kr

Copyright (C) KICHE.all rights reserved.

- Korean Chemical Engineering Research 상단으로