0
# Metrics and Evaluation
1

2
Tools for evaluating model performance including confusion matrices and mean average precision calculations.
3

4
## Capabilities
5

6
### Detection Metrics
7

8
```python { .api }
9
class ConfusionMatrix:
10
    """
11
    Calculate and visualize confusion matrix for classification/detection results.
12
    
13
    Args:
14
        matrix (np.ndarray): Confusion matrix data
15
        classes (list[str]): Class names for labeling
16
    """
17
    
18
class MeanAveragePrecision:
19
    """
20
    Calculate mean Average Precision (mAP) for object detection evaluation.
21
    
22
    Supports COCO-style evaluation with multiple IoU thresholds.
23
    """
24
    
25
    def update(self, predictions: Detections, targets: Detections) -> None:
26
        """Update metric with prediction and ground truth pairs."""
27
        
28
    def compute(self) -> dict:
29
        """Compute final mAP scores and per-class metrics."""
30
```
31

32
## Usage Example
33

34
```python
35
import supervision as sv
36

37
# Initialize metrics
38
confusion_matrix = sv.ConfusionMatrix()
39
map_metric = sv.MeanAveragePrecision()
40

41
# Evaluate predictions against ground truth  
42
for predictions, targets in evaluation_data:
43
    map_metric.update(predictions, targets)
44
    
45
# Get results
46
map_results = map_metric.compute()
47
print(f"mAP@0.5: {map_results['map_50']}")
48
print(f"mAP@0.5:0.95: {map_results['map']}")
49
```