🧠 Project Cere

Multimodal AI for Social Good

AI4Good Lab @Mila • Montreal 2025 Cohort

🔍 Project Overview

Project Cere develops multimodal machine learning models that integrate visual, textual, and audio data to address pressing social challenges. This repository contains our codebase, experiments, and documentation for creating interpretable AI systems with real-world impact.

✨ Features

• Modular Architecture: Easy to extend with new classifiers
• Multiple Classifier Types: Classical ML, Neural Networks, and Ensemble methods
• Flexible Data Pipeline: Support for various fMRI data formats
• Comprehensive Evaluation: Cross-validation, metrics, and visualization
• Hyperparameter Optimization: Built-in grid search capabilities
• Experiment Management: YAML-based configuration system
• Extensible Design: Factory pattern for seamless classifier addition

📋 Table of Contents

• Installation
• Repository Structure
• Features
• Usage
• Project Roadmap
• Team
• License
• Acknowledgments

---

🛠️ Installation

Prerequisites

• Python 3.10+
• Git

Setup

1. Clone the repository:

   git clone https://github.com/marialagakos/AI4Good-MTL-Group-2.git
   cd AI4Good-MTL-Group-2

2. Create and activate virtual environment:

   python -m venv cere-env
   # Linux/MacOS
   source cere-env/bin/activate
   # Windows (PowerShell)
   .\cere-env\Scripts\Activate.ps1

3. Install dependencies:

   pip install --upgrade pip
   pip install -e .  # Editable install for development
   pip install -r requirements.txt  # Optional: Full dependency install

📂 Repository Structure

Project-Cere/
├── main.py                 # Main execution script
├── README.md               # Project documentation
├── data/                   # Raw and processed datasets
│   ├── feature_extraction.py    # Feature extraction utilities
│   ├── DATA_INSTRUCTIONS.md
│   ├── .ipnyb_checkpoints/
│   ├── src/                 
│   │   ├── telepath/
│   │   ├── telepath.egg-info/
│   │   └── temp_audio_chunks/
│   ├── pca_data/
│   ├── loaders.py            # loading (fmri, audio, text, visual)
│   ├── preprocessors.py      # preprocessing
│   ├── transforms.py         # transformations
│   ├── fmri/                    # fMRI data
│   ├── audio/                   # Audio samples
│   ├── transcripts/             # Text corpora
│   └── visual/                  # Image/video data
├── models/                    # Classifier implementations
│   ├── base_classifier.py    # Abstract base class
│   ├── classical/            # Traditional ML methods
│   │   ├── svm.py
│   │   ├── random_forest.py
│   │   └── logistic_regression.py
│   ├── neural/               # Neural network methods
│   │   ├── mlp.py
│   │   ├── cnn.py
│   │   ├── lstm.py
│   │   └── transformer.py
│   └── ensemble/             # Ensemble methods
│       ├── voting.py
│       └── stacking.py
├── utils/                     # Utility functions
│   ├── metrics.py            # Evaluation metrics
│   ├── visualization.py     # Plotting functions
│   └── io_utils.py          # I/O operations
├── experiments/              # Experiment management
│   ├── experiment_runner.py
│   └── hyperparameter_search.py
├── .gitignore              # File control
├── docs/                   # Technical documentation
├── tests/                  # Unit and integration tests
└── LICENSE.md

---

🚀 Usage

Running the Pipeline

python src/main.py --modality all --config configs/default.yaml

Key Arguments

• --modality: Choose audio, text, visual, or all
• --config: Path to YAML configuration file

Jupyter Notebooks

jupyter lab notebooks/

🗺️ Project Roadmap

Phase	Key Deliverables
Data Analysis	EDA reports, preprocessing pipelines
Modeling	Multimodal fusion architectures
Evaluation	Cross-modal attention visualizations
Deployment	Flask API for model serving

👥 Team

• Maria Lagakos - Feature Extraction
• Sophie Strassmann - Creative Director, Pipeline Architecture, and Classification Team
• Yujie Chen - Classification Team
• Keyu Liang - Feature Extraction and Data Migration
• Maria Gallamoso - Feature Extraction Team
• Catherina Medeiros Director of Imaging, Feature Extraction Team

📜 License

This project is licensed under the MIT License - see LICENSE.md for details.

🙏 Acknowledgments

We gratefully acknowledge:

• Jennifer Addison and Yosra Kazemi for their expertise and leadership
• The AI4Good Lab Montreal and Mila team for their support
• Our TA Hugo Berard and Laetitia Constantin

Consulting Scholars and Mentors:

• Rose Landry - Mila
• Adel Halawa - McGill University
• Dr. Lune Bellec - Université de Montréal
• Dr. Mayada Elsabbagh - Transforming Autism Care Consortium
• The Algonauts Project
• Compute Canada for their computational resources
• The Digital Research Alliance of Canada