badoni-sowmya/Anemi-AI

█████╗ ███╗ ██╗███████╗███╗ ███╗██╗ █████╗ ██╗ ██╔══██╗████╗ ██║██╔════╝████╗ ████║██║ ██╔══██╗██║ ███████║██╔██╗ ██║█████╗ ██╔████╔██║██║ ███████║██║ ██╔══██║██║╚██╗██║██╔══╝ ██║╚██╔╝██║██║ ██╔══██║██║ ██║ ██║██║ ╚████║███████╗██║ ╚═╝ ██║██║ ██║ ██║██║ ╚═╝ ╚═╝╚═╝ ╚═══╝╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝ ╚═╝╚═╝ > **Non-invasive anaemia screening and haemoglobin prediction using computer vision and deep learning.** AnemiAI eliminates the need for blood draws in preliminary anaemia screening by analyzing RGB color features extracted from lower eyelid images — making clinical-grade screening accessible in low-resource environments where laboratory infrastructure is unavailable. --- ## The Problem Over 1.6 billion people globally are affected by anaemia. Diagnosis traditionally requires invasive blood collection and laboratory equipment — creating a significant gap in access for rural, remote, and resource-constrained communities. AnemiAI bridges this gap with a camera, a machine learning model, and a web browser. --- ## How It Works 1. User captures an image of their lower eyelid via the web interface 2. The system extracts normalized RGB pixel intensity features from the palpebral conjunctiva 3. A dual-output feedforward neural network runs inference: - **Classifier** → Anaemic / Non-Anaemic - **Regressor** → Estimated haemoglobin (Hb) level in g/dL 4. Results are returned in real time via a REST API --- ## Architecture --- ## Tech Stack | Layer | Technology | |---|---| | Frontend | React.js, Webcam API | | Backend | Python, Flask | | ML Framework | TensorFlow / Keras | | Data Processing | NumPy, Scikit-learn | | Database | MongoDB | --- ## API Reference ### `POST /predict` Accepts normalized RGB percentages extracted from the eyelid region and returns anaemia classification and predicted haemoglobin level. **Request** ```json { "red": 45.2, "green": 29.1, "blue": 25.7 } ``` **Response** ```json { "anaemia": "Yes", "hb": 9.8 …