Average Precision (Metric)

Average Precision (AP) is a metric that summarizes the precision-recall curve into a single value, providing an overall measure of a model’s performance across all classification thresholds. It calculates the area under the precision-recall curve by averaging precision values at different levels of recall.

Mathematically, it is often computed as:

$$\text{AP}=\sum _n({\text{R}}_n-{\text{R}}_{n-1})\cdot {\text{P}}_n$$

Where:

• $({\text{R}}_n)$ and $({\text{R}}_{n-1})$: Recall values at successive thresholds.
• $({\text{P}}_n)$: Precision corresponding to $({\text{R}}_n)$.

Key Point: Higher AP indicates better performance, with a perfect model achieving an AP of 1. It is widely used in object detection and other tasks where precision-recall trade-offs matter.