Intelli_Face_Attendance is an intelligent facial recognition-based attendance system that uses pre-trained models to detect and recognize faces from images and live video feeds, and records attendance by updating a CSV file.
- Face Detection: The system uses the
face_recognitionlibrary to detect faces in images and video frames. It employs the HOG (Histogram of Oriented Gradients) model to locate faces. - Face Recognition: Once faces are detected, the system encodes them and matches these encodings with the pre-trained encodings of known faces to identify individuals.
- Recording Attendance: The system records the presence of recognized individuals by updating a CSV file with their names and the current date.
- Live Video Attendance: The system can capture real-time video from a webcam, detect and recognize faces, and update attendance records dynamically.
- Training Data Preparation: The system processes a dataset of images, encodes the faces, and stores these encodings in a pickle file for quick access during recognition.
- Validation: The system can validate its accuracy by comparing recognized faces against a labeled validation set.
- Display Images: The system can display images with annotated bounding boxes and names, showing which individuals have been recognized.
- Live Feed Display: During live video attendance, the system displays the video feed with real-time annotations of recognized individuals.
The face_recognition library is a simple and powerful library built on top of dlib. It provides easy-to-use functions for face detection, face recognition, and manipulation of images. The library is highly efficient and can work with both images and video streams.
- Face Detection: Detects faces within an image using HOG (Histogram of Oriented Gradients) or CNN (Convolutional Neural Network) models.
- Face Encoding: Converts faces into a high-dimensional vector representation (encoding) that can be used for comparison.
- Face Recognition: Matches a face encoding with a database of known face encodings to recognize the individual.
The HOG (Histogram of Oriented Gradients) model is used for object detection and has been adapted for face detection in the face_recognition library. The HOG method is effective and computationally efficient, making it suitable for real-time applications.
- Gradient Calculation: The image is divided into small connected regions called cells, and for each cell, the gradient direction and magnitude are calculated.
- Histogram Creation: For each cell, a histogram of gradient directions is created. This histogram represents the distribution of gradient directions within the cell.
- Normalization: Histograms are normalized across larger regions of the image, called blocks, to improve accuracy and invariance to lighting changes.
- Feature Vector: The normalized histograms are combined to form a feature vector representing the image.
The HOG model is robust to changes in lighting and pose, making it well-suited for detecting faces in varied conditions.
To use this project, you need to have the following libraries installed:
dlibface_recognitionnumpyopencv-pythonpandasmatplotlib
You can install these dependencies using pip:
pip install dlib face_recognition numpy opencv-python pandas matplotlib