4.1.2. Object Detection

Object detection locates and identifies multiple objects within an image. It answers: "What objects are here and WHERE are they?" Think of it like an annotator drawing boxes around things—you get both the name AND the position.

Key characteristics:

• Output: Bounding boxes with coordinates AND labels
• Can detect multiple objects in one image
• Each detection includes position (x, y, width, height)
• Provides confidence score for each detection

How object detection works:

1. Model scans the image looking for learned objects
2. For each object found, draws a bounding box around it
3. Labels each box with the object category
4. Returns coordinates + labels + confidence scores

Example output:

{
  "objects": [
    {"label": "person", "box": {"x": 100, "y": 50, "w": 80, "h": 200}, "confidence": 0.95},
    {"label": "dog", "box": {"x": 250, "y": 150, "w": 60, "h": 80}, "confidence": 0.88}
  ]
}

Common scenarios:

• Counting products on retail shelves
• Tracking vehicles in traffic monitoring
• Monitoring livestock in agricultural fields
• Safety equipment compliance (hard hats, vests)
• Security systems detecting people or vehicles
• Autonomous vehicles identifying pedestrians and obstacles

Object detection vs. Image classification:

⚠️ Exam Trap: If a question mentions "bounding boxes," "coordinates," "positions," or "locate objects," the answer is object detection—NOT image classification. Keywords like "where," "count," or "track" also signal detection.