What metrics are used for segmentation and object detection?

For segmentation and object detection tasks in computer vision, the following key metrics are used:

• IoU (Intersection over Union) — a measure of overlap between the predicted region and the true object region. Used to evaluate the quality of detection and segmentation.

• mAP (mean Average Precision) — the average precision across different IoU thresholds and classes. Often used to evaluate object detectors.

• Pixel Accuracy — the proportion of correctly classified pixels in the segmentation task.

• Dice Coefficient (F1-score for segmentation) — a measure of similarity between two sets (predicted and true masks), especially popular in medical segmentation.

• Precision and Recall — accuracy and completeness for object detection, important for evaluating detection quality.

Example: for object detection, mAP@0.5 (IoU threshold 0.5) indicates how accurately the model detects objects with at least 50% overlap.