Final Capstone Project Submission - Mold-Kit Web App Student Name: Peris Wangui Supervisor Name: Samiratu Ntohsi Date: December 1, 2025

1. Project Title Mold-Kit: A Machine Learning-Powered Web Application for Mold Detection

2. Project Repository & Demo Links GitHub Repository https://github.com/WanPgui/Final_Product_Capstone Demo Video https://drive.google.com/file/d/10NCNTWTqxIm82Ids1nxYJYpJ9KeC_OtH/view?usp=sharing The repository includes: Complete backend code (FastAPI + TensorFlow model) Frontend UI hosted on Netlify Model artifacts & training notebook Fully updated code after panel feedback Comprehensive README for deployment and installation

3. Project Overview Mold-Kit is a lightweight yet powerful web application designed to detect mold from uploaded images using a trained Machine Learning model. The system enables users to upload a photo, optionally include environmental data (ventilation, humidity, water leaks), and instantly receive an AI-based mold risk assessment. The system aims to help homeowners, renters, property managers, and building inspection teams detect early mold growth to prevent health issues and structural damage. System Components Frontend (Netlify): Interactive and mobile-responsive interface Backend (Render): FastAPI-powered server for inference and API integration AI Model: TensorFlow CNN, optimized for binary mold detection

4. Key Features Drag-and-drop or manual file image upload Optional metadata: location, ventilation, leak status Real-time predictions with confidence levels Color-coded results (green = clean, red = mold detected) No page reload (AJAX async fetch) Automatic image preprocessing and normalization Swagger interactive documentation (/docs) Clean, responsive, and mobile-friendly UI

5. Project Structure Final_Product_Capstone/ ├── Backend/ │ ├── main.py # FastAPI main application │ ├── app.py # Helper functions + preprocessing │ ├── model/ │ │ └── mold_model_final.keras ├── Database/ ├── models/ │ └── mold_Prediction_(1).ipynb ├── static/ # JS, CSS, images ├── templates/ # HTML frontend files └── README.md

6. Installation & Running the Project Backend (FastAPI) cd Backend pip install -r requirements.txt uvicorn main:app --reload Server starts at: http://localhost:8000 Ensure that your model file exists at: Backend/model/mold_model_final.keras Frontend Open index.html in any browser OR Serve using a simple HTTP server: python -m http.server 8001 Make sure the frontend POST request URL matches: http://localhost:8000/predict

7. How Mold-Kit Works (Pipeline) User uploads an image and optionally fills environmental metadata. Frontend sends a POST /predict request using FormData. Backend: Loads and preprocesses the image Resizes to 224x224 Normalizes pixel values Passes tensor into TensorFlow model Model outputs probabilities → converted into label and confidence. Response returned as JSON: { "label": "Mold Detected", "confidence": 0.82, "risk": "High Risk" }

8. API Endpoints & Testing Demo (Swagger / cURL / Postman) Below are fully working demos of all exposed backend routes. 8.1 Swagger UI (Interactive Docs) Navigate to: http://localhost:8000/docs From here, you can: Upload an image Add location/ventilation/leak data Make predictions Validate response formats Test without the frontend 8.2 Test Using cURL POST /predict Upload an image for prediction: curl -X POST "http://localhost:8000/predict"
   -H "accept: application/json"
   -H "Content-Type: multipart/form-data"
   -F "file=@test_image.jpg"
   -F "location=Kitchen"
   -F "ventilation=Poor"
   -F "water_leak=true" Expected Response { "label": "Mold Detected", "confidence": 0.91, "risk": "High Risk" }

GET / Health check endpoint: curl -X GET "http://localhost:8000/" Response: {"status": "API running successfully"} 8.3 Test Using Postman Step-by-step Open Postman → Click New Request Set method to POST Enter URL: http://localhost:8000/predict Go to Body → form-data Add form fields: file (type: File → choose image) location (text) ventilation (text) water_leak (boolean or text) Click Send View JSON prediction result instantly 9. Future Improvements Multi-class mold identification (green/black/white mold) User profiles + prediction history Batch image processing Automated retraining with user-submitted photos PDF report generation with cleaning recommendations Admin dashboard for analytics and trends

10. Tech Stack Component Technology Frontend HTML, CSS, JavaScript Backend Python, FastAPI ML Model TensorFlow, Keras Hosting Netlify (frontend), Render (backend) DB SQLite

11. Updates After Panel Feedback Improved error handling in prediction endpoint Explained preprocessing pipeline clearly in the documentation Updated README to include full setup + deployment steps Reorganized backend directory for clarity Ensured consistent API responses Added endpoint usage examples (cURL, Postman, Swagger)

12. Supervisor & Student Signatures Student: Peris Wangui
    Supervisor: Samiratu Ntohsi Date: 1-12-2025