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
AcneNet: A VGG19-Based Deep Learning Framework for Multi-Class Acne Lesion Detection and Classification Prof.Savita S G1,Bhoomika Reddy2 1Professor,Master of Computer Application,VTU CPGS,Kalaburagi,Karnataka,India 2 Student, Master of Computer Application,VTU CPGS,Kalaburagi, Karnataka,India ---------------------------------------------------------------------***--------------------------------------------------------------------classifying seven distinct skin conditions: papules, Abstract- Acne is a prevalent dermatological condition
blackheads, whiteheads, pustules, nodules, cysts, and normal skin. The system utilizes a custom CNN trained on a curated, high-resolution dataset with diverse skin tones and lighting conditions. Image preprocessing and augmentation improve robustness and generalization. Deployed through a lightweight Flask-based web application, the model delivers real-time classification with visual lesion indicators, bridging clinical diagnostics and consumer-level skin monitoring. This solution offers scalable, objective, and data-driven acne assessment, laying the groundwork for future enhancements such as severity grading, personalized treatment recommendations, and mobile integration.
affecting individuals across all age groups, particularly adolescents and young adults. It manifests in multiple lesion types, including papules, blackheads, whiteheads, nodules, pustules, and cysts, which can adversely affect physical appearance and psychological well-being. Traditional diagnosis relies on manual visual inspection by dermatologists, often leading to subjective and inconsistent assessments. This study proposes AcneNet, a VGG19-based deep learning framework for automated multi-class acne lesion detection and classification. The model is trained on a curated dataset with data augmentation to enhance generalization and reduce overfitting. Integrated into a Flask-based web application, AcneNet delivers real-time, accurate predictions with visual lesion indicators. Experimental evaluation demonstrates high precision, recall, and F1-scores, highlighting its potential in improving dermatological diagnostics.
2. RELATED WORKS Article[1]"A survey on deep learning for skin lesion segmentation" by Z. Mirikharaji et al. in 2023: This survey examines 177 research papers on deep learning methods for skin lesion segmentation, providing comprehensive insights into neural network architectures, datasets, and evaluation metrics. It discusses the strengths and limitations of current segmentation techniques and highlights trends such as the adoption of U-Net and GAN-based models. The survey emphasizes the importance of data augmentation and transfer learning for improving model robustness. It also addresses challenges like class imbalance and image quality variations. The study concludes with future directions for integrating segmentation with classification in dermatology applications.
Keywords: Acne detection, Deep learning, VGG19, Convolutional Neural Network, Multi-class classification, Skin lesions, AI in dermatology, Automated diagnosis.
1.INTRODUCTION Skin health plays a critical role in overall well-being and selfconfidence, particularly among adolescents and young adults. Among dermatological disorders, acne vulgaris is one of the most prevalent and persistent conditions, affecting nearly 85% of individuals aged 12 to 24. Acne manifests in multiple forms, including blackheads, whiteheads, papules, pustules, nodules, and cysts, each necessitating distinct approaches to diagnosis and treatment. While not lifethreatening, acne can significantly impact psychological health, often leading to anxiety, low self-esteem, and depression. Early and precise identification of lesion types is therefore essential for effective intervention and optimal skin management. Traditionally, acne diagnosis depends on manual examination by dermatologists, which, though clinically effective, is susceptible to human error, inconsistencies, and accessibility limitations, particularly in remote or under-resourced areas. The rapid growth of smartphone technology and computational capabilities has created a demand for intelligent, automated acne detection systems that provide fast, reliable, and widely accessible solutions. Leveraging advances in deep learning, especially Convolutional Neural Networks (CNNs), this project proposes an automated, end-to-end system capable of
© 2025, IRJET
|
Impact Factor value: 8.315
Article [2]"Deep Learning in Dermatology: A Systematic Review" by H.K. Jeong et al. in 2022: This review analyzes the application of convolutional neural networks (CNNs) in dermatology, focusing on image-based diagnosis including acne detection. It evaluates algorithm performance across various skin diseases and discusses clinical utility and limitations. The authors review datasets, preprocessing methods, and model architectures, stressing the need for standardized benchmarks. The review also highlights challenges in addressing diverse skin tones and rare lesion types. It ultimately underlines the promise of AI for accessible dermatology while calling for rigorous clinical validation. Article[3] "A novel automatic acne detection and severity quantification scheme based on deep learning" by J. Wang et
|
ISO 9001:2008 Certified Journal
|
Page 414