Weakly Supervised Deep Semantic Segmentation Using CNN and ELM with Semantic Candidate Regions

Table 3

Results on PASCAL VOC 2012 (mIoU in %) for weakly-supervised semantic segmentation with only per-image labels.

PASCAL	background	aeroplane	bicycle	bird	boat	bottle	bus	car	cat	chair	cow	Dining table	dog	horse	motorbike	person	Potted plant	sheep	sofa	train	tv/montir	average
VOC 2012
CCNN [20]	69	26	18	25	20	36	47	47	48	16	38	21	44	35	46	41	30	36	22	39	37	35.3
MIL+ILP+SP-sppxly [21]	77	37	18	25	28	32	42	48	51	13	46	15	51	44	39	38	28	44	20	38	35	36.6
STC [19]	82	63	26	62	28	38	67	63	75	22	53	28	66	58	62	53	33	63	32	45	45	50.7
AE [22]	78	72	29	64	40	58	58	54	63	10	61	36	62	56	63	43	37	65	32	50	39	50.9
Ours	84	68	26	58	47	41	57	67	74	23	73	26	53	57	74	38	43	63	37	40	48	52.2