Table 3. Results for frame-based two-class polyp detection.

	Category	Precision	Recall	F1-score	AP
Faster RCNN	ad	72.8	73.0	72.9	72.9
	hp	42.2	63.1	50.6	42.5
	Mean	57.5	68.1	62.3	57.7
SSD	ad	82.7	77.4	80.0	82.7
	hp	54.6	51.8	53.1	52.5
	Mean	68.6	64.6	66.5	67.6
YOLOv3	ad	89.7	23.2	36.9	61.1
	hp	60.0	16.2	25.5	35.0
	Mean	74.9	19.7	31.2	48.0
RetinaNet	ad	85.4	59.1	69.8	57.9
	hp	52.9	43.7	47.9	40.5
	Mean	69.2	51.4	59.0	49.2
DetNet	ad	73.0	67.5	70.2	60.4
	hp	46.0	65.0	53.8	42.2
	Mean	59.5	66.2	62.7	51.3
RefineDet	ad	92.2	61.3	73.6	81.1
	hp	49.1	86.3	62.6	65.9
	Mean	70.7	73.8	72.2	73.5
YOLOv4	ad	90.5	54.0	67.6	70.4
	hp	54.0	40.6	46.3	42.7
	Mean	72.3	47.3	57.2	56.6
ATSS	ad	79.5	76.3	77.9	80.7
	hp	57.2	68.0	62.2	58.4
	Mean	68.4	72.2	70.2	69.5