🚀 AMS - Advanced Multi-face System

A powerful and modern facial detection and recognition system that enhances RetinaFace's limitations using an innovative 9-segment image technique, multiple face matching libraries, and advanced upscaling methods.

---

🎥 Project Demo

Watch the demo below to see AMS in action:

📸 Overview

AMS (Advanced Multi-face System) is designed to tackle the challenge of detecting and recognizing multiple human faces in images — especially when some are far from the camera or blurred.

• 🔍 Base Detection: Utilizes RetinaFace for initial face detection.
• 🧠 Enhanced Accuracy: Introduces a unique 9-segment image technique that improves detection rates significantly.
• 🧬 Face Matching: Employs multiple face matching libraries like ArcFace, Dlib, Mediapipe, and CV2 for robust identity verification.
• 🧼 Duplicate Removal: Detects and removes duplicate faces across segments with feature comparison.
• 🔍 Face Upscaling: Applies various face unblurring and super-resolution algorithms.

---

🧩 9-Segment Image Technique

To address RetinaFace’s limitations on small or distant faces, AMS splits the image into 9 overlapping blocks and applies detection on each:

🎯 Original Matrix

|  1 |  2 |  3 |  4 |
|  5 |  6 |  7 |  8 |
|  9 | 10 | 11 | 12 |
| 13 | 14 | 15 | 16 |

---

🧠 Segment Blocks

🔹 Block 1

|  1 |  2 |
|  5 |  6 |

🔹 Block 2

|  3 |  4 |
|  7 |  8 |

🔹 Block 3

|  9  | 10 |
| 13 | 14 |

🔹 Block 4

| 11 | 12 |
| 15 | 16 |

🔹 Block 5

|  2 |  3 |
|  6 |  7 |

🔹 Block 6

|  5 |  6 |
|  9 | 10 |

🔹 Block 7

|  7 |  8 |
| 11 | 12 |

🔹 Block 8

| 10 | 11 |
| 14 | 15 |

🔹 Block 9

|  6 |  7 |
| 10 | 11 |

These segmented crops are upscaled back to original size for detection, significantly improving overall accuracy.

---

🧬 Libraries Used

• 🎯 Face Detection: RetinaFace
• 🧑‍🤝‍🧑 Face Matching: ArcFace, Dlib, Mediapipe, CV2
• 🔁 Feature Verification: Facial mesh & feature vector comparison
• 📈 Upscaling: Multiple super-resolution techniques (WIP on glasses!)

---

⚙️ Setup Instructions

🐍 1. Clone the repository

git clone https://github.com/Namitjain07/AMS-Advanced-Multi-face-System.git
cd AMS-Advanced-Multi-face-System

💻 2. Create and activate a new conda environment

conda create -n AMS python=3.10
conda activate AMS

📦 3. Install dependencies

pip install \
retina-face==0.0.17 numpy==1.26.4 gdown==5.2.0 Pillow==11.1.0 \
opencv-python==4.10.0.84 tensorflow==2.17.1 beautifulsoup4==4.12.3 \
filelock==3.17.0 requests[socks]==2.32.3 tqdm==4.67.1 absl-py==1.4.0 \
astunparse==1.6.3 flatbuffers==25.1.21 gast==0.6.0 google-pasta==0.2.0 \
h5py==3.12.1 libclang==18.1.1 ml-dtypes==0.4.1 opt-einsum==3.4.0 \
packaging==24.2 protobuf==4.25.6 setuptools==75.1.0 six==1.17.0 \
termcolor==2.5.0 typing-extensions==4.12.2 wrapt==1.17.2 grpcio==1.70.0 \
tensorboard==2.17.1 keras==3.5.0 tensorflow-io-gcs-filesystem==0.37.1 \
wheel==0.45.1 rich==13.9.4 namex==0.0.8 optree==0.14.0 charset-normalizer==3.4.1 \
idna==3.10 urllib3==2.3.0 certifi==2024.12.14 markdown==3.7 \
tensorboard-data-server==0.7.2 werkzeug==3.1.3 soupsieve==2.6 PySocks==1.7.1 \
MarkupSafe==3.0.2 markdown-it-py==3.0.0 pygments==2.18.0 mdurl==0.1.2 \
mediapipe==0.10.9 pyheif

---

🛠️ Roadmap

• Implement 9-segment strategy
• Add face matching & duplicate removal
• Test and compare upscaling methods
• Improve handling of spectacles
• Integrate with real-time video input
• Build minimal UI for evaluation

---

👨‍💻 Contributions

Want to help? Feel free to open issues or pull requests!

---

📜 License

This project is open-source under the MIT License.

---

💬 Feedback

We're actively improving AMS. If you encounter issues or have suggestions for better face detection (especially for glasses!), open an issue or reach out.

---

🌟 Star this project if you find it helpful!