🎓 Face Recognition Attendance System

A robust, bilingual, real-time face recognition-based attendance system using YOLOv8, DeepFace, TensorFlow, and OpenCV. This project supports both real-time webcam-based tracking and video file-based pipelines with Region of Interest (ROI), logging, frame skipping, and CSV-based attendance.

---

🚀 Project Structure

1. Deep Track Prototype (Webcam-based)

Pipeline:

• Real-time webcam feed.
• Face detection: face_recognition.
• Feature extraction: DeepFace.represent().
• Identity prediction: Pretrained TensorFlow model.
• Attendance CSV generation with name and timestamp.

Highlights:

• Real-time accuracy.
• Simple pipeline to test live tracking.
• CSV output: attendance.csv

---

2. Core 1 (Video + YOLO + DeepFace)

Pipeline:

• Input: Pre-recorded video file.
• ROI (Region of Interest): Defined polygon mask.
• Face detection: YOLOv8 using yolov11l-face.pt.
• Recognition: DeepFace.find() using cosine similarity.
• Output: Annotated video + attendance CSV.

Highlights:

• Supports masked region-based filtering.
• Face matching from pre-encoded database folder.
• Output CSV: attendance_log.csv.

---

3. Core 1 Improved (Optimized)

Pipeline Enhancements:

• Faster detection using yolov11n-face.pt.
• Frame skipping for performance boost.
• Embedding caching to avoid repeated comparisons.
• Overlay of last matched face image and label (“Present”).
• Real-time bounding box display with names.
• Attendance duplication prevention.

Extras:

• Logging with logging module (logs.txt).
• Efficient attendance marking.
• Output files:
  • Annotated video: output.mp4
  • Attendance CSV: final_attendance.csv

---

🧠 Technologies Used

| Module | Purpose | |----------------|------------------------------------| | face_recognition | Real-time face detection | | DeepFace | Face embeddings and identity match | | YOLOv8 | High-speed face detection | | TensorFlow | Custom model for identity prediction (prototype) | | OpenCV | Video I/O, drawing, overlays | | Numpy/Pandas | Data and CSV handling | | Logging | Robust debugging |

---

🗂️ File Structure

├── deep_track_prototype.py       # Webcam-based face tracking
├── core1.py                      # Video + YOLO + DeepFace
├── core1_improved.py             # Optimized video pipeline
├── known_faces/                  # Folder with known images
├── attendance/                   # Output CSV files
├── output/                       # Output annotated video
├── roi_mask.npy                  # ROI polygon mask file
├── logs.txt                      # Runtime logs
├── model.h5                      # Trained TensorFlow model

---

🛠️ Setup Instructions

1. Clone the repo:

   git clone https://github.com/yashbisht077/DeepTrack.git
   

2. Install requirements:

   pip install -r requirements.txt
   

3. Download YOLOv8 weights:

   • Place yolov11l-face.pt and yolov11n-face.pt in the root directory.

4. Run pipelines:

   • Prototype (webcam):

     python detect_faces.py
     

   • Core 1 (video):

     python app.py
     

   • Core 1 Improved:

     python app.py
     

---

📝 Sample Output

• ✅ final_attendance.csv:

Name,Time
Shankar Singh,2025-07-12 09:43:21
Anurag Singh,2025-07-12 09:44:12
...

---

📌 Future Improvements

• Web-based dashboard for CSV management.
• Real-time multi-camera support.
• Automatic ROI calibration using perspective transform.
• Facial spoofing detection.

---

📧 Contact

Shankar Singh
📍 BTech CSE AIML @ GEHU Bhimtal
📧 [shankarbisht1224@gmail.com]

---

📄 License

This project is licensed under the MIT License. Feel free to use, modify, and contribute!