Vision-Based Multiscale Construction Object Detection under Limited Supervision

Table 1

Evaluation results on validation dataset I with different training cases.

Case	Method	mAP
Case 1	FRCD	5.3	16.5	3.3	8.6	3.2	0.7	9.6
	Soft teacher	15.5	34.2	11.7	19.4	11.2	3.6	29.5
	SS-MCOD	16.1	34.7	12.9	21.4	12.8	4.0	30.1
Case 2	FRCD	10.6	31.9	8.8	15.5	6.9	2.5	22.7
	Soft teacher	21.4	45.2	16.4	28.7	14.5	5.3	35.2
	SS-MCOD	22.0	46.1	19.0	28.7	15.0	5.4	36.6
Case 3	FRCD	17.7	36.8	13.5	25.5	12.7	4.8	32.6
	Soft teacher	25.4	50.9	22.5	33.2	19.0	6.0	39.8
	SS-MCOD	28.1	61.3	26.9	36.8	20.0	6.1	41.8
Case 4	FRCD	21.6	45.7	17.1	30.3	16.6	5.7	36.1
	Soft teacher	32.6	59.7	31.7	40.7	24.9	8.5	47.6
	SS-MCOD	35.4	61.3	36.2	44.7	25.1	8.9	49.1

Best evaluation results among three methods.