mAP at IoU Threshold

Problem

Compute average precision for one class at a single IoU threshold.

Sort detections by descending score (ties by index) and match each to its highest-IoU ground-truth box; it is a true positive only when that IoU is at least iou_threshold and the ground truth is still unmatched. detections is a list of [box, score]. AP is the sum of precision at each true positive divided by the number of ground truths, rounded to 4 decimal places; return 0.0 with no ground truths.

Test cases

case 1

in: [[[[0,0,2,2],0.95],[[10,10,12,12],0.9],[[1,1,3,3],0.8]],[[0,0,2,2],[10,10,12,12]],0.5]

expected: 1

case 2

in: [[[[0,0,2,2],0.9],[[0,0,2,2],0.8]],[[0,0,2,2]],0.5]

expected: 1

case 3

in: [[[[0,0,2,2],0.9]],[],0.5]

expected: 0

case 4

in: [[],[[0,0,2,2]],0.5]

expected: 0

Starter codePython

def map_at_iou(detections, gt_boxes, iou_threshold):
    # Your code here
    pass

4 test cases · run your code against them in the editor