제한된 계산 자원 하에서 실시간 계란 결함 검출을 위한 RT-DETR과 YOLOv8의 비교 분석
Comparative Analysis of RT-DETR and YOLOv8 for Real-Time Egg Defect Detection under Computational Constraints
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 논문에서는 실시간으로 계란의 결함을 감지하기 위해 두 가지 고급 물체 감지 모델인 RT-DETR과YOLOv8의 성능을 비교 및 분석한다. 전 세계적으로 계란 수요가 증가함에 따라, 품질과 안전성을보장하는 실시간 결함 감지 기술은 건강 위험과 식품 안전 문제를 예방하기 위해 필수적이다. 이에따라, 두 객체 감지 모델을 이용하여 계란의 투광 이미지를 분석하고, 정상, 균열, 혈반, 탈색, 이물질, 기형 등 6가지 카테고리를 분류한다. 이를 위해, AI-Hub의 11,000개 계란 이미지 데이터 셋을 기반으로훈련, 검증, 평가 비율을 9:1:1로 설정하여 실험을 진행한다. 다음으로, 두 객체 감지 모델의 성능은감지 정확도와 추론 속도 측면에서 평가한다. RT-DETR 모델은 두 가지 버전(r18, r50)으로 학습되며, YOLOv8 모델은 세 가지 버전(nano, medium, large)으로 학습되어 상호 비교된다. 실험 결과, 정확도측면에서는 RT-DETR-r50 모델이 IoU 임곗값 0.5에서 mAP이 0.852(mAP@0.5), 다양한 IoU 임곗값에서0.851(mAP@0.5:0.95)을 도출하였다. 반면, 속도 측면에서는 YOLOv8n 모델이 이미지당 평균 0.0021초의 추론 시간을 기록하여 더 나은 성능을 보였다. 본 연구는 정확성과 속도 간의 균형을 강조하며, 계란 생산 시설에서 효율적인 품질 관리 시스템을 구현하기 위한 유용한 통찰력을 제공한다.
Abstract
In this study, we present a comparative analysis of two state-of-the-art object detection models, RT-DETR and YOLOv8, for real-time egg defect detection. As the global demand for eggs continues to rise, ensuring their quality and safety is paramount for mitigating health risks and addressing food safety concerns. The models were employed to analyze transillumination images of eggs and classify them into six distinct categories: Normal, Crack, Blood Spot, Discoloration, Foreign Body, and Deformation. A dataset comprising 11,000 egg images provided by AI-Hub was utilized, with a train-validation-test split ratio of 9:1:1 to conduct the experiments. The performance of the models was assessed based on two key metrics: detection accuracy and inference speed. For RT-DETR, two variants (r18 and r50) were trained, while YOLOv8 involved the training of three variants (nano, medium, and large). The evaluation results indicated that the RT-DETR-r50 model demonstrated superior accuracy, achieving a mean Average Precision (mAP) of 0.852 (mAP@0.5) at an IoU threshold of 0.5 and 0.851 (mAP@0.5:0.95) across multiple IoU thresholds. Conversely, the YOLOv8n model exhibited exceptional speed, with an average inference time of 0.0021 seconds per image. These findings emphasize the trade-off between accuracy and inference speed, providing critical insights into the development of efficient quality control systems for egg production facilities.
참고문헌 (27)
- P. Mason, "The importance of eggs in an environmentally sustainable diet", Nutrition Bulletin, 48(3), pp.400~410 (2023)[DOI]
- M. Myers, "Eggs: Healthy or Risky? A Review of Evidence from High Quality Studies on Hen’s Eggs", Nutrients, 15(12), pp.2657~ (2023)[DOI]
- S. Réhault-Godbert, "The Golden Egg:Nutritional Value, Bioactivities, and Emerging Benefits for Human Health", Nutrients, 11(3), pp.684~ (2019)[DOI]
- M. Gallo, "Relationships between food and diseases: What to know to ensure food safety", Food Research International, 137, pp.109414~ (2020)[DOI]
- H. Onyeaka, "Preventing chemical contaminants in food: Challenges and prospects for safe and sustainable food production", Food Control, 155, pp.110040~ (2024)[DOI]
- S. Thangaraju, "Food Chemistry: The Role of Additives, Preservatives and Adulteration", pp.189~216 (2021)
- Y. Sun, "The evolution of object detection methods", Engineering Applications of Artificial Intelligence, 133, pp.108458~ (2024)[DOI]
- Y. Zhao, "DETRs Beat YOLOs on Real-time Object Detection", Proc. IEEE, 2024, 16965~16974
- E. L. T. Jun, "A Comparative Analysis of RT-DETR and YOLOv8 for Urban Zone Aerial Object Detection", Proc. 2024 IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS), pp.340~345 (2024)
- X. Yang, "A Computer Vision-Based Automatic System for Egg Grading and Defect Detection", Animals, 13(14), pp.2354~ (2023)[DOI]
- Z. Zhang, "A Real-Time Intelligent Valve Monitoring Approach through Cameras Based on Computer Vision Methods", Sensors, 24(16), pp.5337~ (2024)[DOI]
- M. Turkoğlu, "Defective egg detection based on deep features and Bidirectional Long-Short-Term-Memory", Computers and Electronics in Agriculture, 185, pp.106152~ (2021)[DOI]
- B. Botta, "Eggshell crack detection using deep convolutional neural networks", Journal of Food Engineering, 315, pp.110798~ (2022)[DOI]
- J. Zhang, "Trans4Trans : Efficient Transformer for Transparent Object and Semantic Scene Segmentation in Real-World Navigation Assistance", IEEE Transactions on Intelligent Transportation Systems, 23(10), pp.19173~19186 (2022)[DOI]
- W. Xu, "FEA-Swin: Foreground Enhancement Attention Swin Transformer Network for Accurate UAV-Based Dense Object Detection", Sensors, 22(18), pp.6993~ (2022)[DOI]
- M. Liu, "Bearing-DETR: A Lightweight Deep Learning Model for Bearing Defect Detection Based on RT-DETR", Sensors, 24(13), pp.4262~ (2024)[DOI]
- M. Safaldin, "An Improved YOLOv8 to Detect Moving Objects", IEEE Access, 12, pp.59782~59806 (2024)[DOI]
- M. Bakirci, "Enhancing vehicle detection in intelligent transportation systems via autonomous UAV platform and YOLOv8 integration", Applied Soft Computing, 164, pp.112015~ (2024)[DOI]
- J. Moon, "Object detection under the lens of privacy : A critical survey of methods, challenges, and future directions", ICT Express, 10(5), pp.1124~1144 (2024)[DOI]
- Y. Lee, "A YOLOv8-Based Two-Stage Framework for Non-Destructive Detection of Varroa destructor Infestations in Apis mellifera Colonies", Journal of the Korea Society of Computer and Information, 29(10), pp.137~148 (2024)[DOI]
- H. Kim, "A Real-Time Wildlife Roadkill Detection Approach Using YOLOv8", Journal of Internet of Things and Convergence, 10(5), pp.185~196 (2024)[DOI]
- "Egg Defect Detection Dataset"[URL]
- "Ultralytics YOLOv8: State-of-the-art Object Detection Models"[URL]
- T. Liang, "Pruning and quantization for deep neural network acceleration : A survey", Neurocomputing, 461, pp.370~403 (2021)[DOI]
- A. Gholami, "Low-Power Computer Vision", pp.2021, Chapman and Hall