Abstract
<title>Abstract</title> <p>Traditional Ayurvedic medicine classifies individuals into constitutional types, or doshas, to guide personalized dietary and lifestyle recommendations. This paper presents a mobile framework that combines computer vision and questionnaire-based analysis to provide personalized Ayurvedic wellness recommendations. The proposed system utilizes a YOLO11n object detection model trained on the TMC-Tongue dataset [17] to identify clinically relevant tongue characteristics such as tongue color, coating type, fissures, teeth marks, moisture level, and other visual indicators. These extracted features are used to estimate constitutional tendencies and generate preliminary dosha scores. To improve assessment reliability, tongue-based observations are fused with questionnaire responses describing lifestyle, digestion, sleep quality, hydration status, and energy levels. The combined feature representation is processed through a transparent, rule-based fusion mechanism that estimates dosha imbalances according to predefined Ayurvedic weighting principles. Based on the generated assessment, the system provides personalized Ayurvedic recommendations, including dietary guidance, hydration strategies, and lifestyle modifications. Experimental evaluation demonstrates the effectiveness of the YOLO11n model in learning discriminative tongue features, achieving reliable detection performance across multiple tongue characteristic categories. The proposed framework highlights the potential of integrating artificial intelligence with traditional health assessment approaches for accessible and personalized Ayurvedic wellness monitoring.</p>