Deep Learning (DL) · Advanced level

Advanced level · Module 3: Computer Vision Beyond Classification · Lesson 2 of 4

Evaluating Detection

Measure detection quality with intersection over union and mean average precision.

In this lesson you will

  • Calculate intersection over union
  • Interpret precision, recall and mAP for detection

Key concepts

intersection over union (IoU)true and false detectionsaverage precisionmAP

The lesson

Intersection over union (IoU) measures box overlap: the area shared by the predicted and true boxes divided by their combined area. A detection usually counts as correct when IoU is at least 0.5.

Precision and recall apply to detections: how many predicted boxes were right, and how many real objects were found. Average precision summarises the precision–recall trade-off for one class; mAP averages across classes.

Also check task-level results, such as counting accuracy: a detector with decent mAP may still miscount clustered insects.

Step by step

  1. Calculate IoU for three box pairs by hand.
  2. Compute precision and recall at IoU 0.5.
  3. Compute mAP with a library.
  4. Compare predicted and true counts.

Practice activity

Evaluate the pest detector with IoU, mAP and counting error.

Expected result: Correct metrics and an interpretation linked to the counting task.

Check your understanding

  1. What does an IoU of 1 mean?

    Show answer

    The predicted box exactly matches the true box.

  2. Why check counting accuracy as well as mAP?

    Show answer

    Good box metrics can still hide counting errors, for example in clusters.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.