Cosmetic Skin Type Classification Using CNN With Product Recommendation
Arya Kothari, Dipam Shah, Taksh Soni, Sudhir Dhage
Abstract
The skin spectrum is used in a wide range of studies like dermatology, biometric recognition, cosmetic research and disease detection. In cosmetic research, skin can be classified into four main categories as, normal, dry, oily and combination. The current methods to identify the cosmetic skin type are time consuming and error prone. Recently, Deep Learning algorithms have been in the limelight for various classification problems like text, audio, image and video classifications. In this paper, the applications of Convolutional Neural Network for skin type classification have been studied. To train the model, we have created a dataset of over 80 skin images,collected by web scraping and classified into oily and dry categories. To evaluate the performance of our model, we have used the trained model on a small sample of easily distinguishable images. The results of our CNN classification model show an accuracy of about 85% with a slight bias towards oily images. The results show that Deep Learning has great potential in the field of skin type classification from facial images, and with a dataset of greater size, could give more optimal and even lesser error prone results.