Shishir0315/Anemia-Detection

Anemia Detection using CNN

Anemia Detection using Deep Learning (CNN) & Machine Learning

📌 Project Overview

An end-to-end AI system designed to detect Anemia based on pixel characteristics (% Red, Green, Blue) extracted from nail/conjunctiva images. This project implements and compares two different approaches:

1. Deep Learning: A 1D Convolutional Neural Network (CNN) built with TensorFlow/Keras.
2. Machine Learning: A Random Forest Classifier built with Scikit-Learn.

The system is fully deployed as a local web application using Flask, featuring a modern, user-friendly interface for real-time predictions.

🚀 Key Features

• Dual Model Architecture: Compare CNN vs. Random Forest performance.

• Interactive Web App: Modern UI built with HTML/CSS/JS and Flask backend.

• Data Processing: Automated feature scaling and encoding pipeline.

• Deployment Ready: includes scripts for training, testing, and launching the app.

• src/

  • data_loader.py: Handles loading and preprocessing of Anemia_Dataset.csv.
  • model.py: Defines the 1D CNN architecture.
  • train.py: Main script to train and evaluate the model.

• requirements.txt: List of Python dependencies.

Setup Instructions

1. Prerequisite: Python Environment

Important: Please check the Python environment. If you encounter errors, create a fresh environment:

# Open Anaconda Prompt or Terminal
conda create -n anemia_env python=3.8
conda activate anemia_env
pip install -r requirements.txt

2. Run Training

Do not type python src/train.py (it will fail because python is not in your PATH).

Instead, double-click the run_ml.bat file in this folder, or run this command in your terminal:

.\run_ml.bat

3. Output

The script will:

• Train the model for 50 epochs.
• Save the trained model to anemia_model.h5.
• Save the training accuracy/loss plot to training_history.png.