Face Detection and Recognition

A guide to building an application that can both detect faces in an image and recognize specific individuals.

Libraries and Tools Used

This project leverages a couple of powerful Python libraries for these tasks:

• face_recognition: A popular and easy-to-use library that wraps around `dlib` for a simple API.
• cv2 (OpenCV): For loading and displaying images, and drawing bounding boxes.

1. Load and Encode Known Faces

• What the code does: It loads images of known individuals, finds their faces, and computes their 128-dimensional face encodings. These encodings are a unique numerical representation of a face.
• Behind the scenes: The `face_recognition` library uses a pre-trained deep learning model to generate these encodings. The process is a form of feature extraction where the model learns the most important features of a face.
• Abstraction: The `face_recognition.face_encodings` function abstracts the complex process of running a neural network to get these embeddings.

# pip install face_recognition opencv-python
import cv2
import face_recognition

# Load images of known people and get their encodings
known_face_encodings = []
known_face_names = []

# Example: Load an image of a known person
known_image = face_recognition.load_image_file("known_person.jpg")
known_encoding = face_recognition.face_encodings(known_image)[0]
known_face_encodings.append(known_encoding)
known_face_names.append("Known Person")

print(f"Loaded and encoded {len(known_face_names)} known faces.")

2. Detect and Recognize Faces in a New Image

• What the code does: It loads a new image, finds all the faces in it, and computes their encodings. It then compares these new encodings to the list of known encodings.
• Behind the scenes: The comparison uses a simple distance metric (like Euclidean distance) to find the closest match between the new face encoding and the known face encodings. A low distance means a high similarity.
• Abstraction: The `face_recognition.compare_faces` and `face_recognition.face_distance` functions abstract the entire comparison process, allowing us to easily determine if a face matches a known one.

# Load a test image
test_image = face_recognition.load_image_file("test_image.jpg")
face_locations = face_recognition.face_locations(test_image)
face_encodings = face_recognition.face_encodings(test_image, face_locations)

for face_encoding in face_encodings:
    # Compare the new face to the known faces
    matches = face_recognition.compare_faces(known_face_encodings, face_encoding)
    name = "Unknown"

    # Find the best match
    face_distances = face_recognition.face_distance(known_face_encodings, face_encoding)
    best_match_index = np.argmin(face_distances)
    if matches[best_match_index]:
        name = known_face_names[best_match_index]
    print(f"Detected face: {name}")

3. Visualize the Results

• What the code does: It iterates through the detected face locations and draws a bounding box and the recognized name around each face using OpenCV.
• Behind the scenes: OpenCV provides simple functions like `cv2.rectangle` and `cv2.putText` to draw on images. This allows for a clear visual representation of the detection and recognition results.
• Abstraction: The OpenCV drawing functions abstract the low-level pixel manipulation needed to add graphical elements to an image.

# Convert test image from RGB to BGR for OpenCV
rgb_test_image = cv2.cvtColor(test_image, cv2.COLOR_RGB2BGR)

for (top, right, bottom, left), face_encoding in zip(face_locations, face_encodings):
    matches = face_recognition.compare_faces(known_face_encodings, face_encoding)
    name = "Unknown"
    # Find the best match and assign name
    face_distances = face_recognition.face_distance(known_face_encodings, face_encoding)
    best_match_index = np.argmin(face_distances)
    if matches[best_match_index]:
        name = known_face_names[best_match_index]

    # Draw a box and name label
    cv2.rectangle(rgb_test_image, (left, top), (right, bottom), (0, 255, 0), 2)
    cv2.rectangle(rgb_test_image, (left, bottom - 35), (right, bottom), (0, 255, 0), cv2.FILLED)
    font = cv2.FONT_HERSHEY_DUPLEX
    cv2.putText(rgb_test_image, name, (left + 6, bottom - 6), font, 1.0, (255, 255, 255), 1)

# Display the resulting image
cv2.imshow('Face Recognition', rgb_test_image)
cv2.waitKey(0)
cv2.destroyAllWindows()