인터넷정보학회논문지 (Journal of Internet Computing and Services)

  • 제25권5호
  • /
  • Pages.95-105
  • /
  • 2024
  • /
  • 1598-0170(pISSN)
  • /
  • 2287-1136(eISSN)
한국인터넷정보학회 (Korean Society for Internet Information)
권호 표지
DOI QR코드

DOI QR Code

모델 크기별 데이터 증강 비율 탐구를 통한 YOLO 기반 의류 이미지 다중 카테고리 분류 연구

Exploring Data Augmentation Ratios for YOLO-Based Multi-Category Clothing Image Classification by Model Size

  • Seyeon Park (Yonsei University Graduate School of Information) ;
  • Sunga Hwang (Yonsei University Graduate School of Information) ;
  • Beakcheol Jang (Yonsei University Graduate School of Information)
  • 투고 : 2024.07.13
  • 심사 : 2024.08.30
  • 발행 : 2024.10.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

최근 여러 의류 쇼핑 플랫폼 및 의류 관련 산업에서 AI를 도입하여 소비자의 니즈를 충족시키고 구매력을 높이는 체계를 도입함에 따라, 의류의 카테고리와 색상을 정확히 분류하는 필요성이 급증하고 있다. 본 연구는 구매자 리뷰 이미지를 사용하여 한 이미지내 여러 카테고리의 다양한 의류와 해당 색상을 분류하는 딥러닝 모델을 개발함으로써 이와 같은 문제를 해결하고자 한다. 구매자 리뷰 이미지 데이터를 직접 크롤링하여 데이터 증강 등 다양한 전처리 과정을 거친 후, YOLOv10 모델을 이용하여 의류의 객체를 탐지하고 이를 카테고리별로 분류한다. 이후, 이미지의 색상을 더 잘 추출하기 위해 고려한 크롭 방법을 통해 의류 영역을 자르고, 색상 차트와의 유사도를 계산하여 가장 유사한 색상명을 추출하는 방법을 구현한다. 실험 결과, 본 연구의 접근 방식이 효과적임을 보여주며, 모델 크기 및 증강 배율이 높을수록 성능이 향상함을 확인하였다. 사용한 모델은 의류 카테고리 및 색상 추출에서 모두 안정적인 성능을 기록하였으며, 그 신뢰성을 입증하였다. 제안된 시스템은 사용자 리뷰 이미지를 기반으로 한 정확한 의류 카테고리와 색상 분류를 통해 고객 만족도 및 구매력을 향상할 뿐만 아니라, 자동화된 패션 분석에 대한 추가 연구의 기초를 마련한다. 또한, 패션 트렌드 분석, 재고 관리, 마케팅 전략 수립 등 관련 산업의 여러 분야에도 활용될 수 있는 확장성을 지닌다.

With the recent adoption of AI by various clothing shopping platforms and related industries to meet consumer needs and enhance purchasing power, the necessity for accurate classification of clothing categories and colors has surged. This paper aims to address this issue by developing a deep learning model that classifies various clothing items and their colors within a single image using buyer review images. After directly crawling buyer review image data and performing various preprocessing steps such as data augmentation, we utilized the YOLOv10 model to detect clothing objects and classify them into categories. Subsequently, to improve color extraction, we implemented a cropping method to isolate clothing regions in the images and calculated the similarity with a color chart to extract the most similar color names. Our experimental results show that our approach is effective, with performance increasing with model size and augmentation scale. The employed model showed stable performance in both clothing category and color extraction, proving its reliability. The proposed system not only enhances customer satisfaction and purchasing power by accurately classifying clothing categories and colors based on user review images but also lays the foundation for further research in automated fashion analysis. Moreover, it possesses the scalability to be utilized in various fields of the related industry, such as fashion trend analysis, inventory management, and marketing strategy development.

키워드

과제정보

This work was supported by the National Research Foundation of Korea (NRF) funded by Korean Government under Grant RS-2023-00273751.

참고문헌

  1. Shajini, Majuran, and Amirthalingam Ramanan, "An improved landmark-driven and spatial-channel attentive convolutional neural network for fashion clothes classification," The Visual Computer, Vol. 37, No. 6, pp. 1517-1526, 2021. https://doi.org/10.1007/s00371-020-01885-7
  2. Hye-Suk Kim, "Classification and Combination of Fashion Items Using CNN-Based Deep Learning," Journal of Digital Contents Society, Vol. 22, No. 3, pp. 475-482, 2021. http://dx.doi.org/10.9728/dcs.2021.22.3.475
  3. JW Yang, "Analysis of consumer purchasing behavior models according to contents of fashion shopping platform applications : Focused on the top three women's clothing companies users' rankings," Master's Thesis, Hongik University, 2022. https://dcoll.hongik.ac.kr/srch/srchDetail/000000027932
  4. Cychnerski, Jan, et al., "Clothes detection and classification using convolutional neural networks," 2017 22nd IEEE international conference on emerging technologies and factory automation (ETFA), IEEE, 2017. https://doi.org/10.1109/ETFA.2017.8247638
  5. Abd Alaziz, Hadeer M., et al., "Enhancing Fashion Classification with Vision Transformer (ViT) and Developing Recommendation Fashion Systems Using DINOVA2," Electronics, Vol. 12, No. 20, pp. 4263, 2023. https://doi.org/10.3390/electronics12204263
  6. Sivaranjani, Lingala, et al., "Fashion Recommendation System Using Machine Learning," 2023 4th International Conference on Smart Electronics and Communication (ICOSEC), IEEE, 2023. https://doi.org/10.1109/ICOSEC58147.2023.10275967
  7. Nocentini, Olivia, et al., "Image classification using multiple convolutional neural networks on the fashion-MNIST dataset," Sensors, Vol. 22, No. 23, pp. 9544, 2022. https://doi.org/10.3390/s22239544
  8. JH Cho, HS Kwon, Yoon KH, "Perspectives of 'Ubiquitous Health Care System' for Diabetes Management," Diabetes & Metabolism Journal, Vol. 30, No. 2, pp. 87-95, 2006. http://dx.doi.org/10.4093/jkda.2006.30.2.87
  9. Dosovitskiy, Alexey, et al., "An image is worth 16x16 words: Transformers for image recognition at scale," arXiv, 2020. https://doi.org/10.48550/arXiv.2010.11929
  10. Xiao, Han, Kashif Rasul, and Roland Vollgraf, "Fashionmnist: a novel image dataset for benchmarking machine learning algorithms," arXiv, 2017. https://doi.org/10.48550/arXiv.1708.07747
  11. Mohanty, D. K. et al., "Modified Convolutional Neural Network for Fashion Classification," 2022 International Conference on Advancements in Smart, Secure and Intelligent Computing (ASSIC), pp.1-9, 2022. https://doi.org/10.1109/ASSIC55218.2022.10088358
  12. HS Jeong, SY Lee, CK Lee, "Deep learning-based clothing attribute classification using fashion image data," Smart Media Journal, Vol. 13, No. 4, pp. 57-64, 2024. http://doi.org/10.30693/SMJ.2024.13.4.57
  13. He, Kaiming, et al., "Deep residual learning for image recognition," Proceedings of the IEEE conference on computer vision and pattern recognition(CVPR), 2016. https://doi.org/10.48550/arXiv.1512.03385
  14. Tan, Mingxing, and Quoc Le, "Efficientnet: Rethinking model scaling for convolutional neural networks," International conference on machine learning(ICML), PMLR, pp. 6105-6114, 2019. https://doi.org/10.48550/arXiv.1905.11946
  15. Guo, Calvin H., "Search My Favorites by Color: Fashion Parsing through Color Classification," CS230: Deep Learning, Winter 2020, Stanford University, CA, 2020.
  16. He, Kaiming, et al., "Mask r-cnn," Proceedings of the IEEE international conference on computer vision (ICCV), pp. 2961-2969, 2017. https://doi.org/10.48550/arXiv.1703.06870
  17. Girshick, Ross, et al., "Rich feature hierarchies for accurate object detection and semantic segmentation," Proceedings of the IEEE conference on computer vision and pattern recognition(CVPR), pp. 580-587, 2014. https://doi.org/10.48550/arXiv.1311.2524
  18. Girshick, Ross, "Fast r-cnn," Proceedings of the IEEE international conference on computer vision(ICCV), pp. 1440-1448, 2015. https://doi.org/10.48550/arXiv.1504.08083
  19. Ren, Shaoqing, et al., "Faster r-cnn: Towards real-time object detection with region proposal networks," Advances in neural information processing systems (NeurIPS), 2015. https://doi.org/10.48550/arXiv.1506.01497
  20. Liu, Wei, et al., "Ssd: Single shot multibox detector," Computer Vision-ECCV 2016: 14th European Conference, Part I 14. Springer International Publishing, 2016. https://doi.org/10.48550/arXiv.1512.02325
  21. Lin, Tsung-Yi, et al., "Focal loss for dense object detection," Proceedings of the IEEE international conference on computer vision(ICCV), pp. 2980-2988, 2017. https://doi.org/10.48550/arXiv.1708.02002
  22. Zhang, Shifeng, et al., "Single-shot refinement neural network for object detection," Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), pp. 4203-4212, 2018. https://doi.org/10.48550/arXiv.1711.06897
  23. Redmon, Joseph, et al., "You only look once: Unified, real-time object detection," Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), pp. 779-788, 2016. https://doi.org/10.48550/arXiv.1506.02640
  24. Wang, Ao, et al., "Yolov10: Real-time end-to-end object detection," arXiv, 2024. https://doi.org/10.48550/arXiv.2405.14458