• Version
• Download 21
• File Size 458.57 KB
• File Count 1
• Create Date 6 December 2025
• Last Updated 6 December 2025

AI-Based Skin Anomaly Detection and Recommendation

 

 

Dr Poornima Raikar1,Prajwal Itagi 2, Shrilakshmi Mulagundmath3

1Department of Computer Science(AIML),KLS Vishwanathrao Deshpande Institute Of Technology Haliyal,India 2Department of Computer Science(AIML),KLS Vishwanathrao Deshpande Institute Of Technology Haliyal,India 3Department of Computer Science(AIML),KLS Vishwanathrao Deshpande Institute Of Technology Haliyal,India

 

 

Abstract - The expanding demand for accessible machine learning solutions in academic and research environments necessitates innovative web-based platforms for model training and evaluation. This AI-powered ML training web application revolutionizes data analysis workflows by integrating PyTorch deep learning models with an intuitive Flask backend and interactive frontend interfaces. The system addresses critical challenges in ML experimentation, including dataset preparation, model training, result visualization, and user feedback collection through seamless web deployment and real-time processing capabilities. Built on robust web frameworks, the platform enables multiple users to upload datasets, initiate training sessions, monitor progress via live logs, and access comprehensive results without local infrastructure dependencies. Key features include automated dataset validation, PyTorch model training with configurable hyperparameters, dynamic result dashboards, and integrated questionnaires for performance assessment, ensuring reproducible experiments and streamlined collaboration. The application leverages large-scale neural networks for tasks like classification or regression, delivering intelligent predictions with syntax-aware error handling and autosave mechanisms that prevent session disruptions. Interactive components such as upload interfaces, training monitors, and result visualizations empower users with real-time insights, reducing setup time and enhancing model interpretability. AI-driven automation handles preprocessing pipelines, hyperparameter optimization suggestions, and performance benchmarking, significantly lowering barriers for non-expert users while maintaining research-grade accuracy. The extensible architecture supports multiple data formats and model architectures through modular Python scripts and responsive HTML/CSS/JS interfaces, facilitating efficient project management and scalability. This platform combines web accessibility with advanced ML capabilities, inspired by modern training pipelines and deployment tools, fostering innovation in AI experimentation and enabling teams to achieve superior model performance through collaborative, browser-based intelligence.

 

Key Words:Skin anomaly detection, MobileNetV2, deep learning, skincare recommendation, web application, computer vision.

Download