{"id":25952417,"url":"https://github.com/junayed-hasan/smile-recognition-fusion","last_synced_at":"2026-06-07T11:32:02.010Z","repository":{"id":268799627,"uuid":"905497679","full_name":"junayed-hasan/smile-recognition-fusion","owner":"junayed-hasan","description":"An automatic, efficient, and effective framework for spontaneous smile classification combining handcrafted D-Marker features with transformer-based automatic features using diverse feature fusion techniques. Achieves state-of-the-art results across benchmark smile datasets.","archived":false,"fork":false,"pushed_at":"2024-12-19T01:29:23.000Z","size":97,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2024-12-19T02:22:45.575Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/junayed-hasan.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-12-19T00:27:21.000Z","updated_at":"2024-12-19T01:29:27.000Z","dependencies_parsed_at":"2024-12-19T02:22:50.790Z","dependency_job_id":"6243ebce-faf3-4be5-8df6-2c43fc386b29","html_url":"https://github.com/junayed-hasan/smile-recognition-fusion","commit_stats":null,"previous_names":["junayed-hasan/smile-recognition-fusion"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/junayed-hasan%2Fsmile-recognition-fusion","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/junayed-hasan%2Fsmile-recognition-fusion/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/junayed-hasan%2Fsmile-recognition-fusion/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/junayed-hasan%2Fsmile-recognition-fusion/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/junayed-hasan","download_url":"https://codeload.github.com/junayed-hasan/smile-recognition-fusion/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":241868475,"owners_count":20033822,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2025-03-04T14:51:12.106Z","updated_at":"2026-06-07T11:32:02.004Z","avatar_url":"https://github.com/junayed-hasan.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# HadaSmileNet: Hadamard Fusion for Facial Emotion Recognition\n\n[![Paper](https://img.shields.io/badge/Paper-ICDM%202025-blue)](link-to-paper-when-available)\n[![License](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE)\n[![Python](https://img.shields.io/badge/Python-3.8-blue.svg)](https://python.org)\n\n![Architecture Diagram](fusion_archi.png)\n\n## 🎉 News\n- **August 2025**: Paper accepted at IEEE International Conference on Data Mining (ICDM) 2025\n- **September 2025**: Repository made public for camera-ready submission\n- **Conference Date**: November 12-14, 2025\n\n## Table of Contents\n1. [Abstract](#abstract)\n2. [Key Contributions](#key-contributions)\n3. [Results](#results)\n4. [Repository Structure](#repository-structure)\n5. [Getting Started](#getting-started)\n6. [Reproducing Results](#reproducing-results)\n7. [Citation](#citation)\n8. [License](#license)\n\n---\n\n## Abstract\n\n**HadaSmileNet** introduces a novel feature fusion framework that directly integrates transformer-based representations with physiologically-grounded D-Markers through parameter-free multiplicative interactions for genuine smile recognition. Unlike multi-task learning approaches that require auxiliary task supervision and complex loss balancing, our method achieves optimal performance through direct Hadamard multiplicative fusion while maintaining computational efficiency during inference.\n\n**Key Achievements:**\n- 🏆 **State-of-the-art results** on 4 benchmark datasets\n- ⚡ **26% parameter reduction** compared to multi-task alternatives  \n- 🎯 **Statistical significance** with p \u003c 0.001 across all improvements\n- 📊 **99.7% accuracy** on MMI, **100% accuracy** on BBC dataset\n\n![Performance Comparison](performance_progression.png)\n\n---\n\n## Key Contributions\n\n1. **Novel Fusion Architecture**: Direct integration of handcrafted D-Marker features with transformer representations through Hadamard multiplicative fusion\n2. **Computational Efficiency**: 26% parameter reduction and simplified training compared to multi-task learning frameworks  \n3. **Comprehensive Evaluation**: Systematic comparison of 15 fusion strategies across 4 benchmark datasets\n4. **Statistical Validation**: Rigorous significance testing demonstrating reliable performance improvements\n5. **Practical Deployment**: Inference-time efficiency suitable for real-time applications\n\n---\n\n## Results\n\n### Performance on Benchmark Datasets\n\n| Dataset   | Accuracy | Improvement | Statistical Significance |\n|-----------|----------|-------------|-------------------------|\n| UvA-NEMO  | **88.7%** | +0.8%      | p \u003c 0.001***           |\n| MMI       | **99.7%** | -           | -                       |\n| SPOS      | **98.5%** | +0.7%      | p \u003c 0.001***           |\n| BBC       | **100%**  | +5.0%      | p \u003c 0.001***           |\n\n### Computational Efficiency\n- **Training Time Reduction**: 42.3% (p \u003c 0.001)\n- **Parameter Reduction**: 26.0% vs DeepMarkerNet\n- **Inference Speed**: Comparable to baseline methods\n\n---\n\n## Repository Structure\n\n```\nHadaSmileNet/\n├── Code/\n│   ├── BBC/                    # BBC dataset experiments\n│   │   ├── [15 fusion method notebooks]\n│   │   ├── dataset/           # Model checkpoints\n│   │   ├── core/             # Utility scripts\n│   │   ├── labels/           # Cross-validation splits\n│   │   └── npy/             # Dataset numpy files\n│   ├── SPOS/                  # SPOS dataset experiments\n│   ├── UvA-NEMO/             # UvA-NEMO dataset experiments  \n│   └── MMI/                   # MMI dataset experiments\n├── figures/                   # Paper figures and plots\n├── statistical_analysis/     # Statistical significance tests\n├── requirements.txt\n├── LICENSE\n└── README.md\n```\n\n---\n\n## Getting Started\n\n### Prerequisites\n- Python 3.8+\n- CUDA-compatible GPU (recommended)\n- 8GB+ RAM\n\n### 1. Clone the Repository\n```bash\ngit clone https://github.com/junayed-hasan/HadaSmileNet.git\ncd HadaSmileNet\n```\n\n### 2. Install Dependencies\n```bash\npip install -r requirements.txt\n```\n\n### 3. Download Datasets\nEach dataset folder contains download instructions in the `npy/` directory. Place the downloaded numpy files in the corresponding dataset folders.\n\n### 4. Verify Installation\n```bash\ncd Code/BBC\njupyter notebook concatenation.ipynb  # Test with simple fusion method\n```\n\n---\n\n## Reproducing Results\n\n### Cross-Validation Setup\n- **UvA-NEMO**: 10-fold cross-validation\n- **BBC**: 10-fold cross-validation  \n- **MMI**: 9-fold cross-validation\n- **SPOS**: 7-fold cross-validation\n\n### Running Experiments\n\n1. **Individual Fusion Methods**:\n   ```bash\n   cd Code/[DATASET_NAME]\n   jupyter notebook [fusion_method].ipynb\n   ```\n\n2. **All 15 Fusion Strategies**:\n   Navigate to each dataset folder and run all notebooks sequentially.\n\n3. **Statistical Analysis**:\n   ```bash\n   cd statistical_analysis\n   python statistical_significance_test.py\n   ```\n\n### Expected Runtime\n- **Single experiment**: ~2-4 hours per dataset\n- **Complete evaluation**: ~8-12 hours for all fusion methods\n- **Hardware**: NVIDIA T4 GPU or equivalent\n\n---\n\n## Citation\n\nIf you use this work in your research, please cite:\n\n```bibtex\n@inproceedings{hasan2025hadasmilenet,\n  title={HadaSmileNet: Hadamard fusion of handcrafted and deep-learning features for enhancing facial emotion recognition of genuine smiles},\n  author={Hasan, Mohammad Junayed and Mohammed, Nabeel and Rahman, Shafin},\n  booktitle={Proceedings of the IEEE International Conference on Data Mining (ICDM)},\n  year={2025},\n  organization={IEEE}\n}\n```\n\n---\n\n## Acknowledgments\n\n- IEEE International Conference on Data Mining (ICDM) 2025\n- Johns Hopkins University Computer Science Department\n- North South University Apurba NSU R\u0026D Lab\n\n---\n\n## License\n\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n---\n\n## Contact\n\n**Mohammad Junayed Hasan**  \nComputer Science Department, Johns Hopkins University  \n📧 junayedhasan100@gmail.com \n🔗 [GitHub](https://github.com/junayed-hasan)\n\n---\n\n**© 2025 Mohammad Junayed Hasan. Published at IEEE ICDM 2025.**\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjunayed-hasan%2Fsmile-recognition-fusion","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjunayed-hasan%2Fsmile-recognition-fusion","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjunayed-hasan%2Fsmile-recognition-fusion/lists"}