研究生: |
黃文雄 Wen-Shiung Huang |
---|---|
論文名稱: |
基於機器學習之雲端皮膚頭皮檢測系統 A Cloud-based Intelligent Skin & Scalp Analysis System |
指導教授: |
花凱龍
Kai-Lung Hua |
口試委員: |
鄭文皇
孫士韋 陳永耀 陳宜惠 |
學位類別: |
博士 Doctor |
系所名稱: |
電資學院 - 資訊工程系 Department of Computer Science and Information Engineering |
論文出版年: | 2022 |
畢業學年度: | 110 |
語文別: | 中文 |
論文頁數: | 103 |
中文關鍵詞: | 深度學習 、YOLOv5 、頭皮檢測 、皮膚檢測 、頭髮密度測量 、頭髮髮徑 |
外文關鍵詞: | scalp detection, hair density measurement, hair diameter |
相關次數: | 點閱:312 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
隨著物聯網與人工智慧及雲端運算的高速發展,傳統笨重的PC皮膚頭皮檢測儀器系統已經無法滿足消費者對於隨時、隨地獲取自身皮膚頭皮資訊的需求,因此結合移動裝置與具有人工智能的雲端皮膚頭皮檢測與分析系統,市場前景看好。另外近年來,深度學習已應用於臨床醫學影像分析中,有優異的表現。隨著這些方法的持續發展,具有深度學習功能的產品已廣泛應用在美容美髮與醫美大健康行業中。
本文以皮膚和頭皮檢測為研究標的,開發了一套以深度學習檢測算法的雲端皮膚和頭皮檢測系統,提供12項檢測項目,並對美髮和醫美需求的頭髮密度測量和頭髮髮徑測量,做了深度學習算法的研究和實現。本文的具體研究內容如下:
1.完成一個可以兩千台以上裝置同時運作,基於雲端的智能頭皮及皮膚檢測系統,系統已實際應用在2000家以上醫美和沙龍門店,並且建立了一個超過幾百萬張頭皮和皮膚數據資料庫。
2.進行以美髮沙龍比超商密集的場域, 藉由 AI檢測和體驗延伸消費商機。置入社群互動(Line,FB,Whatsapp)推送檢測報告、智慧推薦、分潤商城、支付等服務,讓消費者從服務體驗到產品導購。解決店家美容產品庫存和銷售的問題,讓傳統服務場域轉型為新時代的最佳零售場域!
3. 頭髮密度測量(HDM)和頭髮髮徑 ,常被用來判斷頭髮及頭皮健康程度。我們完成一個長短髮狀況都可以辦別,以偏光鏡頭取樣照片,可以避免頭皮和毛髮油垢產生影像干擾,可以直接標定毛囊和分辨1-3株毛髮建構一個髮密度檢測算法。我們以實驗證明修改yolov5-our演算法,比更先進的yolov7有更高的準確率(mAP@50 77.1 Vs 73.8) ,由Yolov5s改進增加了5.2%的準確率(mAP@50 71.9-->77.1)。
With the rapid development of the Internet of Things, artificial intelligence and cloud computing, the traditional bulky PC skin and scalp detection instrument system has been unable to meet the needs of consumers to obtain their own skin and scalp information anytime and anywhere. Therefore, the skin and scalp detection and analysis system which combines mobile devices and cloud-based artificial intelligence has a promising market prospect. In addition, in recent years, deep learning has been applied to clinical medical image analysis with excellent performance. With the continuous development of these methods, products with deep learning functions have been widely used in the beauty ,hairdressing ,aesthetic medicine and health industries.
Taking skin and scalp detection as the research object, this paper develops a cloud-based skin and scalp detection system based on deep learning detection algorithms, provides 12 detection items, and measures hair density and diameter for hairdressing and aesthetic medicine needs. Research and implementation of deep learning algorithms. The specific research contents of this paper are as follows:
1. Complete a cloud-based intelligent scalp and skin detection system that can operate more than 2,000 devices at the same time. The system has been practically applied to more than 2,000 beauty shops and hair salons, and a database of more than several million scalp and skin data has been established.
2. Carry out areas where hair salons are denser than Convenience stores, and extend consumption opportunities through AI detection and experience. Incorporate community interaction (Line, FB, Whatsapp) send test report, intelligent recommendation, profit-sharing mall, payment ,and other services, allowing consumers to experience from service to product shopping guide. Solve the problem of inventory and sales of beauty products in stores, and transform traditional service areas into the best retail areas in the new era!
3. Hair Density Measurement (HDM) and hair diameter are often used to judge the health of hair and scalp. We can distinguish between long and short hair conditions. Using a polarized lens to sample photos can avoid image interference caused by scalp and hair grease, and can directly calibrate hair follicles and distinguish 1-3 hairs to construct a hair density detection algorithm. We experimentally prove that the modified yolov5-our algorithm has higher accuracy than the more advanced yolov7 (mAP@50 77.1 Vs 73.8), and the Yolov5s improvement increases the accuracy by 5.2% (mAP@50 71.9-->77.1 ).
[1] Shih, H.C. An unsupervised hair segmentation and counting system in microscopy images. IEEE Sens. J. 2014, 15, 3565–3572. [CrossRef]
[2] Shih, H. A precise automatic system for the hair assessment in hair-care diagnosis applications. Skin Res. Technol. 2015, 21, 500–507. [CrossRef] [PubMed]
[3] Zhang, Q.; Eun, S. Design and implementation of an automatic hair counting system. J. Digit. Art Eng. Multimed. 2014, 1, 75–86.
[4] Kim, W.; Kim, H.; Rew, J.; Hwang, E. A hair density measuring scheme using smartphone. In Proceedings of the Korea Information Processing Society Conference, Seoul, Korea, 10–12 June 2015; pp. 1416–1419.
[5] Seo S, Park J. Trichoscopy of alopecia areata: hair loss feature extraction and computation using grid line selection and eigenvalue[J]. Computational and Mathematical Methods in Medicine, 2020, 2020: 6908018.
[6] Kim H, Kim W, Rew J, et al. Evaluation of hair and scalp condition based on microscopy image analysis[C]. 2017 International Conference on Platform Technology and Service (PlatCon), 2017: 1-4.
[7] Chang, W.J.; Chen, L.B.; Chen, M.C.; Chiu, Y.C.; Lin, J.Y. ScalpEye: A deep learning-based scalp hair inspection and diagnosis system for scalp health. IEEE Access 2020, 8, 134826–134837. [CrossRef]
[8] Chen, Ming-Che, et al. "A Hair Volume Analysis System Using Deep Learning and Image Processing Technologies." 2021 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW). IEEE, 2021.
[9] Jakubík, J. Dataset Enhancement in Hair Follicle Detection: ESENSEI Challenge. Proc. FedCSIS 2018, 19–22. Available online: https://annals-csis.org/proceedings/2018/drp/388.html (accessed on 23 November 2021). [CrossRef]
[10] Gallucci, A.; Znamenskiy, D.; Pezzotti, N.; Petkovic, M. Hair counting with deep learning. In Proceedings of the BIA, Varna, Bulgaria, 24–27 September 2020; pp. 5–9.
[11] Sacha J P, Caterino T L, Fisher B K, et al. Development and qualification of a machine learning algorithm for automated hair counting[J]. International Journal of Cosmetic Science, 2021, 43: S34-S41.
[12] Urban G, Feil N, Csuka E, et al. Combining deep learning with optical coherence tomography imaging to determine scalp hair and follicle counts[J]. Lasers in Surgery and Medicine, 2021, 53(1): 171-178.
[13] Kim M, Kang S, Lee B-D. Evaluation of Automated Measurement of Hair Density Using Deep Neural Networks[J]. Sensors, 2022, 22(2): 650.
[14] Wen-Shiung Huang, Bing-Kai Hong, Wen-Huang Cheng, Shih-Wei Sun, and Kai-Lung Hua. "A Cloud-based Intelligent Skin and Scalp Analysis System," In Proceedings of the IEEE International Conference on Visual Communications and Image Processing (VCIP), Taichung, Taiwan, Dec. 9 - 12, 2018.
[15] X Wu, D Sahoo, SCH Hoi. Recent advances in deep learning for object detection Neurocomputing 396 (2020) 39–64 41
[16] D. G. Lowe, Object recognition from local scale-invariant features, in:ICCV, 1999.
[17] R. Lienhart, J. Maydt, An extended set of haar-like features for rapid object detection, in: International Conference on Image Processing, 2002.
[18] N. Dalal, B. Triggs, Histograms of oriented gradients for human detection,in: CVPR, 2005.
[19] H. Bay, T. Tuytelaars, L. Van Gool, Surf: Speeded up robust features,in: ECCV, 2006.
[20] M. A. Hearst, S. T. Dumais, E. Osuna, J. Platt, B. Scholkopf, Support vector machines, in: IEEE Intelligent Systems and their applications,1998.
[21] D. Opitz, R. Maclin, Popular ensemble methods: An empirical study, in:Journal of artificial intelligence research, 1999.
[22] P. Viola, M. J. Jones, Robust real-time face detection, in: IJCV, 2004.
[23] Y. Freund, R. E. Schapire, et al., Experiments with a new boosting algorithm,in: ICML, 1996.
[24] P. F. Felzenszwalb, R. B. Girshick, D. McAllester, D. Ramanan, Object detection with discriminatively trained part-based models, in: TPAMI,2010.
[25] A. Krizhevsky, I. Sutskever, G. E. Hinton, Imagenet classification with deep convolutional neural networks, in: NeurIPS, 2012.
[26] K. Fukushima, S. Miyake, Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition, in: Competition and cooperation in neural nets, 1982.
[27] Y. LeCun, L. Bottou, Y. Bengio, P. Haffner, Gradient-based learning applied to document recognition, in: Proceedings of the IEEE, 1998.
[28] Uijlings,RrJ, Sande V D, et al. Selective search for object recognition[]J. International Journal of Computer Vision ,2013,104(2):154-171.
[29] Ren, S., He, K.,Girshick, R., Sun, J. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149.
[30] Sermanet, Pierre, et al. OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks[J], Eprint Arxiv,2013:1312.6229.
[31] Naqvi, Syed Faraz, et al. Real-Time Stress Assessment Using Sliding Window Based Convolutional Neural Network[JJ. Sensors, 2020, 20(16):4400.
[32] Girshick R, Donahue J, Darrell T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation[C]//IEEE Computer Society, 2013:580-587.
[33] He K, Zhang X, Ren S, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[JJ. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015,37(9):1904-1916.
[34] Girshick,R.Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015:1440-1448.
[35] Redmon J, Divvala S, Girshick R, et al. You Only Look Once: Unified, Real-Time Object Detection[J. Computer Vision and Pattern Recognition, 2016:779-788.
[36] Redmon J, Farhadi A. YOLO9000: Better, Faster, Stronger[C]// IEEE Conference on Computer Vision and Pattern Recognition,2017:6517-6525.
[37] Wang S, Li A, Chen J, et al. RSnet: An improvement for arknet[J]. Eprint Arxiv,2020:2002.03729.
[38] Redmon J, Farhadi A. YOLOv3: An Incremental Improvement[J]. Eprint Arxiv,2018:1804.02767.
[39] 許騰,唐貴進,劉清萍,鮑秉坤.基於空洞卷積和FocalLoss的改進YOLOv3演算法門門,南京郵電大學學報(自然科學版),2020,40(06):100-108.
[40] Bochkovskiy A, Wang C Y, Liao H. YOLOv4: Optimal Speed and Accuracy of Object Detection[J. Eprint Arxiv, 2020:2004.10934.
[41] Mahto P, Garg P, Seth P, et al. Refining Yolov4 for Vehicle Detection[]J. Intemnational Journal of Advanced Research in Engineering and Technology, 2020,11(5).
[42] Jocher G. Yolovs[EB/OL].[2020-08-10], https://github.com/ultralytics/yolovs.
[43] Liu W, Anguelov D, Erhan D, et al. SSD: Single Shot MultiBox Detector[C]// European Conference on Computer Vision, 2016:21-37.
[44] Jeong J, Park H, Kwak N. Enhancement of SSD by concatenating feature maps for object detection[C]// Eprint Arxiv, 2017:1705.09587.
[45] Dermatoscopy: Physics and principles B Nirmal - Indian Journal of Dermatopathology and Diagnostic …, 2017
[46] Ashok, Sahana. "Dermatoscopy: A New Diagnostic Approach for Lesions on Mucous Membrane." (2022).
[47] Benjumea, Aduen, et al. "YOLO-Z: Improving small object detection in YOLOv5 for autonomous vehicles." arXiv preprint arXiv:2112.11798 (2021).
[48] Kim, Ildoo, Younghoon Kim, and Sungwoong Kim. "Learning loss for test-time augmentation." Advances in Neural Information Processing Systems 33 (2020): 4163-4174.