Anemia is a major global health concern affecting well-being and productivity. Traditional diagnostic methods, while effective, are often invasive, costly, and inaccessible in resource-limited settings. This paper presents an automated, non-invasive anemia detection system using conjunctiva image analysis, integrating advanced machine learning and deep learning techniques for accurate and accessible diagnosis. A deep learning-based object detection model precisely segments the conjunctiva, extracting key regions for analysis. RGB intensity values from the segmented regions serve as input features for a machine learning model, which predicts hemoglobin levels and classifies anemia severity into Non-anemic, Mild, Moderate, and Severe, incorporating demographic factors such as age and gender. To enhance reliability and clinical applicability, the system includes visualization techniques that highlight key predictive features. Optimizations such as improved bounding box handling, standardized image processing, and robust error management ensure high performance and consistency. A user-friendly web interface enables medical professionals to upload images, receive predictions, and analyze results for informed decision-making. By automating anemia detection, this system reduces dependence on invasive testing and facilitates early diagnosis, particularly in underserved regions. Combining deep learning with interpretable AI, the framework offers a scalable and practical solution to improve global anemia diagnostics.