Table 3.

Total number of correct detections (true positives; TP), incorrect detections (false positives; FP), missed detections (false negatives; FN), and the average precision, recall and F1 of the circuli detector and of a second labeller when compared to labeller one (ground truth) across 87 transects.

	Circuli detector	Human labeller
TP	5944 (66; 47–94)	6060 (66; 49–100)
FP	617 (5; 1–16)	546 (4; 0–16)
FN	375 (2; 0–13)	259 (1; 0–12)
Precision	0.90 (0.93; 0.78–0.98)	0.95 (0.95; 0.77–1)
Recall	0.94 (0.97; 0.84–1)	0.96 (0.98; 0.89–1)
F1	0.92 (0.94; 0.82–0.98)	0.94 (0.96; 0.84–0.99)

Median values of TP, FP, FN, and F1 per transect are provided in brackets, followed by the 5th and 95th percentile.