International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
SKINSENSE AI: A DEEP LEARNING–DRIVEN MODEL FOR DERMATOLOGICAL ASSESSMENT USING COMPUTER VISION Mrs. S. Cynthia Juliet1, S. M. Sindhu2 1Assistant Professor & Head, Department of Computer Applications, Jaya College of Arts & Science, Chennai. 2PG Student, Department of Computer Applications, Jaya College of Arts & Science, Chennai.
-----------------------------------------------------------------------***--------------------------------------------------------------------
the gap between sophisticated dermatology tools and daily skincare requirements.
Abstract - Skin analysis is an important process in dermatology and cosmetology because it plays a significant role in diagnosing various skin conditions and identifying suitable skincare routines. Dermatologists' subjective and inconsistent visual judgment is a major component of traditional skin examination techniques. Automated skin analysis has drawn a lot of attention with the development of AI and computer vision because it offers a dependable, effective, and scalable solution. In this paper, a deep learning- driven skin analysis system that uses facial images to predict skin type and identify common skin issues is proposed. To identify several skin conditions at once, the system makes use of preprocessing, skin segmentation, convolutional neural networks, and multi-label classification. Dermatology clinics, cosmetic companies, and consumer applications can benefit from the suggested solution's increased dependability, decreased manual labor, and real-time analysis.
2. LITERATURE REVIEW [1]. Many studies have used computer vision techniques to analyze skin images for detecting texture, color changes, and visible skin issues. Earlier research used basic image processing methods like filtering and segmentation to identify problems such as redness or acne. [2]. Recent works show that deep learning models, especially CNNs, can predict skin diseases more accurately than traditional methods. Researchers used networks like VGG, ResNet, and MobileNet to classify skin conditions from facial images. [3]. Some studies focused on building mobile applications that allow users to take a photo and get diagnostic results. These systems use lightweight deep learning models to identify issues like acne, oiliness, dryness, and pigmentation in real-time.
Keywords: Computer Vision, Skin Type Classification, Skin Problem Prediction, Deep Learning, Image Processing, CNN Models, Dermatology AI, Feature Extraction, Mobile Skin Analysis, Health Monitoring System
[4]. Researchers have used datasets such as DermNet, HAM10000, and self-collected skin image datasets. Many papers highlight that lack of diverse data (different skin tones, lighting conditions) is a major challenge for accurate prediction.
1. INTRODUCTION Determining the right skincare treatments requires understanding skin types and identifying skin issues. But manual skin diagnosis takes a lot of time and expertise. People frequently misjudge their own skin type, which leads to inappropriate product use that can exacerbate skin conditions. Automated skin analysis using computer vision has become a feasible and affordable solution with the growing availability of high-resolution mobile cameras. Skin texture, color variations, oiliness, pore interpretability, this paper presents a workable, scalable method for developing customer behavior prediction systems that support both batch and near-real-time use cases. By applying deep learning models, it is possible to detect subtle differences that might go unnoticed by the human eye. The combination of CNNs and transfer learning allows machines to learn complex patterns from thousands of facial images. With the help of a straightforward mobile photo, users will be able to obtain an immediate and customized skin assessment, bridging
© 2025, IRJET
|
Impact Factor value: 8.315
3. METHODOLOGY Data collection, preprocessing, feature extraction, model training, and evaluation are some of the stages in the methodology. For training, both custom face images and images from public datasets are utilized. To guarantee consistency, lighting correction, gamma adjustment, filtration, and facial alignment are all part of the preprocessing step. The face region is detected using MTCNN. Deep CNN models are responsible for feature extraction. Models pre-trained on massive datasets like ImageNet, such as VGG16, MobileNetV2, and ResNet50, are used to apply transfer learning. From edges and colors to intricate facial textures, these models extract multi-level features. Skin type classification is performed with a softmax activation layer that assigns one of the predefined types. For skin problem prediction, a sigmoid-based multi-label classifier identifies the presence of multiple conditions.
|
ISO 9001:2008 Certified Journal
|
Page 1068