基於深度學習之人臉膚質偵測系統

Human Facial Skin Defects Detection System using Deep Learning

林若瑋(Jo-Wei Lin)

指導教授 ： 秦群立

中山醫學大學/健康管理學院/醫學資訊學系/碩士(2020年)

摘要

人臉膚質狀況探究一直是大家會關注的議題，民眾會為了了解與改善自己的臉部膚質狀況而尋求醫學美容診所的專業幫助。近年來深度學習的興起使得許多領域的學者也逐漸開始將深度學習應用於各個領域中尋求新的突破，在瀏覽許多文獻之後，我們發現將深度學習應用於膚質偵測的相關應用稀少。為了能夠讓人以較為方便的方式了解到自身的臉部膚質狀況，本論文透過訓練Mask R-CNN深度學習演算法模型，讓模型學習在人臉影像上標記膚質狀況，其中Mask R-CNN在實例分割上取得了很好的成果，其優點使本論文採用Mask R-CNN達成本論文的目標，同時本論文亦使用Cycle-GAN之深度學習演算法進行膚質偵測的實驗，將標記後的膚質偵測影像視為另外一種風格的影像進行轉換，在經過數次的實驗，比較其結果顯示Mask R-CNN在完成人臉膚質偵測與標記的任務能夠與Cycle-GAN相較之下獲得較佳的結果。未來亦能將使結果真正落實於伊美相關的場域。

關鍵字

臉部膚質偵測 ； 深度學習 ； 實例分割 ； Cycle-GAN ； Mask R-CNN

並列摘要

Human facial skin has always been an issue that everyone will pay attention to, and many people will seek professional help from medical aesthetic clinics in order to understand and improve their skin condition. In recent years, the rise of deep learning has led scholars in many fields to gradually apply deep learning in various fields to find new breakthroughs. After browsing many documents, we find that the application of deep learning in skin detection is rare. In order to let people understand their own skin condition in a more convenient way, this paper apply Mask R-CNN, which is a deep learning algorithm model, to let the model learn to mark the skin conditions on human face images, and we have very good results by using Mask R-CNN. Mask R-CNN has its advantages on semantic segmentation, which is good to reaching our goal of this paper. At the same time, we also use Cycle-GAN, which is also a deep learning algorithm, for skin detection experiments in this paper. The facial skin detection images are regarded as another style of image conversion. After several experiments, the comparison results show that compares to Cycle-GAN, Mask R-CNN can perform better in completing the task of detecting and marking the skin texture of the human face.

並列關鍵字

Facial skin detection ； deep learning ； semantic segmentation ； Cycle-GAN ； Mask R-CNN

參考文獻

[1] S. Sunder, “Relevant Topical Skin Care Products for Prevention and Treatment of Aging Skin,” Facial Plastic Surgery Clinics of North America, 2019.

[2] Jennifer Burris MS, RD William Rietkerk MD, MBA, Kathleen Woolf PhD, RD, FACSM, “Acne: The Role of Medical Nutrition Therapy,” Journal of the Academy of Nutrition and Dietetics, Vol. 113, pp. 416-430, 2013.

[3] J.W. Marson, H.E. Baldwin, “An Overview of Acne Therapy, Part 2: Hormonal Therapy and Isotretinoin,” Dermatologic Clinics, Vol. 37, pp. 195-203, 2019.

[4] C. Y. Chang, S. C. Li, P. C. Chung, J. Y. Kuo, and Y. C. Tu, “Automatic facial skin defect detection system,” Proc. IEEE International Conference on Broadband, Wireless Computing, Communication and Applications, pp. 527-532, 2010.

[5] S. Chatterjee, D. Dey, S. Munshi, S. Gorai, “Extraction of features from cross correlation in space and frequency domains for classification of skin lesions,” Biomedical Signal Processing and Control, Vol. 53, pp. , 2019