3D Model Retrieval System

This project involves the development of a 3D model indexing and retrieval system based on shape descriptors. The system is designed to calculate shape descriptors for a query 3D model and measure its similarity to models in a database, returning the closest matches.

Features

• Shape Descriptors: Implementation of 3D Fourier coefficients and Zernike moments for compact and invariant representations of 3D models.
• 3D Model Normalization: Ensures invariance to rotation, translation, and scaling.
• Similarity Search: Measures similarity using descriptors and returns the most similar models.
• Mesh Simplification: Evaluates performance with and without mesh reduction.
• Web Application: Built using Flask (backend) and React with Next.js (frontend).

Technologies Used

• Python: Implementation of algorithms using libraries like NumPy and SciPy.
• Flask: Backend for the web application.
• MongoDB: Storage for 3D model data.
• React & Next.js: Frontend for a dynamic and performant user interface.

How It Works

1. Upload Models: Upload 3D models to the system, specifying their category, thumbnail, and path.
2. Shape Descriptor Calculation:
   • Compute 3D Fourier coefficients and Zernike moments for the uploaded models.
   • Normalize models to ensure invariance to rotation, translation, and scaling.
3. Similarity Search:
   • Select a query model and perform a similarity search.
   • Choose between searches with or without mesh reduction.
4. Results Display:
   • The system returns similar models with details such as name, category, similarity score, distance, and reduction type.

Key Algorithms

3D Fourier Coefficients

• Represents the structure of 3D models in frequency space.
• Steps:
  1. Normalize the 3D mesh (center and scale).
  2. Apply the 3D Fourier transform.
  3. Extract magnitudes of Fourier coefficients for invariance.

Zernike Moments

• Utilizes orthogonal Zernike polynomials for shape representation.
• Steps:
  1. Convert Cartesian coordinates to spherical coordinates.
  2. Compute Zernike polynomials for the 3D model.
  3. Normalize moments for invariance.

Results and Analysis

• Fourier Coefficients: Effective for compact representation and capturing model structure.
• Zernike Moments: Highly precise for complex shapes.
• Mesh Reduction: Slight drop in performance with reduced meshes, but descriptors remain robust.

Installation and Usage

1. Clone the repository:

   git clone https://github.com/amine-sabbahi/content-based-3d-search.git
   cd content-based-3d-search

2. Start the backend server:

   cd back-end
   python app.python

3. Start the frontend server:

   cd front-end
   npm run install
   npm run dev

4. Access the application at http://localhost:3000.

Contributors

• Mohamed Amine Sabbahi.

• Fatima Zahra Arichi.

• Aymane Mahri.

---

Abdelmalek Essaadi University Faculty of Sciences and Techniques

• Department : Computer Engineering
• Master : AI & DS
• Module : Multimedia Mining and Indexing
• Framed by : Pr. M’hamed AIT KBIR