Intermediate

Object Detection with OpenCV

Detect faces, objects, and features using classical and deep learning approaches built into OpenCV.

Haar Cascade Classifiers

Haar cascades are pre-trained XML classifiers that detect objects by scanning an image at multiple scales. OpenCV ships with cascades for faces, eyes, smiles, and more.

Python — Face detection with Haar cascades
import cv2

# Load pre-trained cascade
face_cascade = cv2.CascadeClassifier(
    cv2.data.haarcascades + "haarcascade_frontalface_default.xml"
)
eye_cascade = cv2.CascadeClassifier(
    cv2.data.haarcascades + "haarcascade_eye.xml"
)

img = cv2.imread("group_photo.jpg")
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# Detect faces
faces = face_cascade.detectMultiScale(gray, 1.3, 5)
for (x, y, w, h) in faces:
    cv2.rectangle(img, (x, y), (x+w, y+h), (255, 0, 0), 2)

    # Detect eyes within each face region
    roi_gray = gray[y:y+h, x:x+w]
    eyes = eye_cascade.detectMultiScale(roi_gray)
    for (ex, ey, ew, eh) in eyes:
        cv2.circle(img, (x+ex+ew//2, y+ey+eh//2), ew//2, (0, 255, 0), 2)

print(f"Found {len(faces)} faces")

Template Matching

Python — Find a template in an image
img = cv2.imread("screenshot.png")
template = cv2.imread("icon.png")
h, w = template.shape[:2]

# Match template
result = cv2.matchTemplate(img, template, cv2.TM_CCOEFF_NORMED)
threshold = 0.8
locations = np.where(result >= threshold)

# Draw rectangles around matches
for pt in zip(*locations[::-1]):
    cv2.rectangle(img, pt, (pt[0]+w, pt[1]+h), (0, 255, 0), 2)

Contour Detection

Python — Find and analyze contours
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)

contours, hierarchy = cv2.findContours(
    thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE
)

for cnt in contours:
    area = cv2.contourArea(cnt)
    if area > 500:  # Filter small contours
        x, y, w, h = cv2.boundingRect(cnt)
        cv2.rectangle(img, (x, y), (x+w, y+h), (0, 255, 0), 2)

        # Shape approximation
        epsilon = 0.02 * cv2.arcLength(cnt, True)
        approx = cv2.approxPolyDP(cnt, epsilon, True)
        print(f"Vertices: {len(approx)}, Area: {area}")

DNN Module for Deep Learning

OpenCV's DNN module can load and run pre-trained models from TensorFlow, PyTorch, ONNX, Caffe, and Darknet (YOLO).

Python — YOLO object detection
# Load YOLO model
net = cv2.dnn.readNetFromDarknet("yolov4.cfg", "yolov4.weights")
net.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)

# Prepare input
blob = cv2.dnn.blobFromImage(img, 1/255.0, (416, 416),
                              swapRB=True, crop=False)
net.setInput(blob)

# Get output layer names
ln = net.getLayerNames()
ln = [ln[i - 1] for i in net.getUnconnectedOutLayers()]

# Forward pass
outputs = net.forward(ln)

# Process detections
for output in outputs:
    for detection in output:
        scores = detection[5:]
        class_id = np.argmax(scores)
        confidence = scores[class_id]
        if confidence > 0.5:
            # Scale bounding box to image size
            box = detection[0:4] * np.array([w, h, w, h])
            (cx, cy, bw, bh) = box.astype("int")
            x = int(cx - bw / 2)
            y = int(cy - bh / 2)
            cv2.rectangle(img, (x, y), (x+int(bw), y+int(bh)),
                         (0, 255, 0), 2)
✅
Modern alternative: For production object detection, consider using OpenCV's DNN module with ONNX models exported from PyTorch or TensorFlow. This gives you the speed of OpenCV's optimized inference with the accuracy of modern architectures like YOLOv8.