Multiscale Meets Spatial Awareness: An Efficient Attention Guidance Network for Human Parsing

Table 4

Per-label comparison of per-class IoU with eight state-of-the-art methods on the LIP validation set.

Method	SegNet	FCN-8s	DeepLab (VGG16)	Attention	DRN50 + Vortex	DeepLab (Res-101)	SS-JPPNet	SS-NAN	Baseline (VGG16)	AG-Net (VGG16)	AG-Net (Dense-121)
Hat	26.60	39.79	57.94	58.87	57.50	59.76	59.75	63.86	58.06	63.27	66.24
Hair	44.01	58.96	66.11	66.78	67.37	66.22	67.25	70.12	68.14	69.52	71.25
Gloves	0.01	5.32	28.50	23.32	13.55	28.76	28.95	30.63	12.50	23.92	38.74
Sunglasses	0.00	3.08	18.40	19.48	18.82	23.91	21.57	23.92	18.96	29.59	30.45
U-clothes	34.46	49.08	60.94	63.20	62.52	64.95	65.30	70.27	62.33	64.62	67.12
Dress	0.00	12.36	23.17	29.63	23.88	33.68	29.49	33.51	22.49	26.37	31.06
Coat	15.97	26.82	47.03	49.70	48.51	52.86	51.92	56.75	45.12	47.31	53.56
Socks	3.59	15.66	34.51	35.23	33.64	37.67	38.52	40.18	30.72	41.02	45.69
Pants	33.56	49.41	64.00	66.04	65.79	68.05	68.02	72.19	64.98	68.98	73.34
Jumpsuit	0.01	6.48	22.38	24.73	14.10	26.15	24.48	27.68	15.66	20.52	25.52
Scarf	0.00	0.00	14.29	12.84	0.84	17.44	14.92	16.98	8.05	11.75	17.22
Skirt	0.00	2.16	18.74	20.41	19.41	25.23	24.32	26.41	12.31	22.65	27.60
Face	52.38	62.65	69.70	70.58	71.99	70.00	71.01	75.33	71.92	73.40	74.34
L-arm	15.30	29.78	49.44	50.17	52.17	50.42	52.64	55.24	51.57	58.03	62.06
R-arm	24.23	36.63	51.66	54.03	55.25	53.89	55.79	58.93	54.82	60.35	63.98
L-leg	13.82	28.12	37.49	38.35	40.23	39.36	40.23	44.01	38.03	46.35	48.75
R-leg	13.17	26.05	34.60	37.70	37.63	38.27	38.80	41.87	37.85	45.18	48.30
L-shoe	9.26	17.76	28.22	26.20	27.58	26.95	28.08	29.15	25.67	32.88	38.64
R-shoe	6.47	17.70	22.41	27.09	26.53	28.36	29.03	32.64	25.13	35.03	39.20