🌟🔥 Ultimate Face Recognition System 🔥🌟

High-performance. Real‑time. Beautifully engineered. Designed for creators who want power and style.

A complete face recognition toolkit built with InsightFace and OpenCV, including three modules:

• Basic face recognition from stored photos
• Semi‑Advanced face recognition with manual/automatic photo addition
• Advanced long‑distance, smooth, CPU‑optimized real‑time recognition

This README summarizes all three scripts and how to use them.

---

📂 File Overview

---

🗂️ Modules At a Glance

✨ Three scripts, three power levels.

---

1. advance.py – Long‑Distance + Smooth Real‑Time Recognition

Features:

• CPU‑optimized for stability
• Long‑range face detection
• Ultra‑smooth 60 FPS camera processing
• Dynamic text scaling
• Auto‑save embeddings

2. semiadvance.py – Auto‑Add Photos + Multi‑Source Input

Features:

• Automatically converts dropped photos to embeddings
• Supports webcam, IP camera, and MP4
• Live saving with keypress
• Auto face indexing

3. basic.py – Photo Imports + Simple Recognition

Features:

• Loads photos from a folder
• Extracts embeddings automatically
• Real‑time recognition from webcam

---

🚀 Installation

---

🔧 Quick Setup

---

Install dependencies:

pip install insightface opencv-python numpy

If using GPU (optional):

pip install onnxruntime-gpu

Create needed folders:

mkdir saved_faces
mkdir photos

---

▶️ Running the Scripts

---

🎬 Choose Your Mode

---

Basic Mode

python basic.py

• Add photos inside photos/
• The script auto‑detects and saves faces

---

Semi‑Advanced Mode

python semiadvance.py

Menu options:

• 1 → Webcam
• 2 → IP Camera
• 3 → MP4 File

Drop images directly into saved_faces/ to auto‑add people.

---

Advanced Mode

python advance.py

Controls:

• S → Save current face
• Q → Quit

This mode is ideal for:

• Long‑distance recognition
• Crowd scanning
• Fast real‑time processing

---

🧠 How Recognition Works

---

🧩 Behind the Magic

---

1. InsightFace detects the face
2. A 512‑D embedding vector is generated
3. The embedding is compared with saved .npz files
4. If the distance < threshold (0.68–0.75), the identity is shown

---

📸 Saving New Faces

Method A (Camera)

Press S or N depending on script.

Method B (Drop Photos)

Just place .jpg/.png images in:

saved_faces/

The system converts them on next run.

Method C (Photos Folder)

Drop photos into:

photos/

The basic.py script will process them.

---

🎨 UI Features

---

💎 What You See

---

• Dynamic text scaling based on face size
• FPS counter
• Clear color‑coded labels
• Auto‑reloading camera if disconnected

---

⚙️ Recommended Threshold Values

---

🎯 Tuning for Accuracy

---

• 0.68 → Accurate long‑range recognition
• 0.75 → Flexible real‑world lighting

---

📌 Folder Structure

---

🗃️ Project Layout

---

project/
│── advance.py
│── semiadvance.py
│── basic.py
│── saved_faces/
│   └── person_name.npz
│── photos/
│   └── image.jpg
│── README.md  ← this file

---

🙌 Credits

---

❤️ Contributors & Tech

---

• InsightFace for face detection & embeddings
• OpenCV for real‑time video processing

---

📞 Support

Need improvements, optimizations, or a GUI version? Reach out using the contact details below.

• Email: kadamamit462@gmail.com
• Instagram: @piyush_kadam96k