blood grouping detection using image processing

Act as an expert Machine Learning Engineer and Python developer. I need a complete, end-to-end project solution for detecting blood grouping (e.g., ABO system) from input images using image processing and deep learning techniques.

The final output must be a fully functional, runnable project that can be deployed as a simple API or a mini-website.

Please structure the response to include the following components:

1. Data Preparation & Model Training: Specify the necessary steps for data collection (or assume a standard dataset structure if real data is unavailable) and the architecture for the image classification/segmentation model (e.g., using TensorFlow/PyTorch).
2. Image Processing Pipeline: Detailed code for loading, pre-processing (resizing, normalization), and running inference on an input image.
3. API/Web Application Layer: Implement a backend using a Python web framework (e.g., Flask or FastAPI) to expose an endpoint that accepts an image file and returns the predicted blood group result.
4. Complete Code: Provide all necessary Python scripts, configuration files (e.g., requirements.txt), and clear setup instructions so that the solution is immediately runnable.

Specific Requirements & Constraints:

• Technology Stack: Must be implemented entirely in Python, utilizing libraries such as OpenCV, NumPy, and a deep learning framework (e.g., CNNs).
• Output Format: The API endpoint must return the blood group prediction in a clear, structured format (e.g., JSON).
• Clarity: All code segments must be well-commented and logically structured.
• Focus: The primary goal is a deployable prototype demonstrating the end-to-end process from image input to result output.