Glass Refraction Distortion Object Detection via Abstract Features

Table 3

Results of different algorithms for detection of glass surface interference images in GRI dataset (black font indicates the best data, underlined font indicates second best).

Algorithms	Target occlusion		Water drop		Stains		Blurry		Average
	Acc(%)	IoU	Acc(%)	IoU	Acc(%)	IoU	Acc(%)	IoU	Acc(%)	IoU
Deep face	84.36	0.78	82.71	0.82	83.95	0.79	81.49	0.81	83.13	0.80
VGG face	88.54	0.79	85.62	0.75	81.37	0.73	82.05	0.78	84.40	0.76
TBE-CNN	90.71	0.90	87.38	0.86	85.29	0.82	83.72	0.85	86.78	0.86
DA-GAN	93.52	0.84	91.26	0.85	87.48	0.83	89.41	0.81	90.42	0.83
PEN-3D	86.74	0.88	85.29	0.92	82.74	0.87	83.18	0.86	84.49	0.88
LMZMPM	94.15	0.87	89.47	0.91	91.83	0.86	90.62	0.89	91.52	0.88
Ours	94.31	0.89	92.42	0.91	91.71	0.90	92.27	0.91	92.68	0.90