Skip to main content

Table 1 COCO metrics [42]

From: Comparative analysis of deep learning image detection algorithms

Average precision (AP):   
AP % AP at IoU .50:.05:.95 (primary challenge metric)
AP(IoU = .50) % AP at IoU .50 (PASCAL VOC metric)
AP(IoU = .75) % AP at IoU .75 (strict metric)
AP across scales   
AP (small) % AP for small objects Area < 322
AP (medium) % AP for medium objects 322 < area < 962
AP (large) % AP for large objects Area > 962
Average Recall (AR)  
AR (max=1) % AR given 1 detection per image
AR (max=10) % AR given 10 detections per image
AR (max=100) % AR given 100 detections per image
AR across scales   
AR (small) % AR for small objects Area < 322
AR (medium) % AR for medium objects 322 < area < 962
AR (large) % A for large objects Area > 962