RoadSegNet: a deep learning framework for autonomous urban road detection

Journal of Engineering and Applied Science

Table 4 Compared classification performance metrics for each network

S. no.	Metrics	Right road			Left road			Environment
S. no.	Metrics	ResNet50	XceptionNet	MobileNet-V3	ResNet50	XceptionNet	MobileNet-V3	ResNet50	XceptionNet	MobileNet-V3
Training dataset
1	Accuracy	94.96	97.77	99.15	76.17	86.24	99.46	98.36	97.40	97.50
2	IoU	89.53	92.41	94.29	57.87	41.84	49.33	97.08	96.78	97.34
3	Mean BF score	0.8008	0.8093	0.8679	0.7185	0.6631	0.7256	0.8675	0.8443	0.8502
Testing dataset
1	Accuracy	92.7	96.52	95.94	31.32	52.79	68.10	97.91	97.28	97.07
2	IoU	85.07	90.23	89.29	26.81	33.06	40.82	95.20	95.67	95.73
3	Mean BF score	0.7640	0.7794	0.8080	0.5115	0.5689	0.6331	0.8479	0.8279	0.8253
Validation dataset
1	Accuracy	91.42	94.61	95.73	71.38	90.40	93.82	97.97	97.74	97.46
2	IoU	84.99	89.39	90.07	47.78	37.04	36.39	95.94	96.54	96.51
3	Mean BF score	0.7585	0.7917	0.8251	0.6823	0.6151	0.6774	0.8559	0.8524	0.8449