Robust object detection for autonomous driving based on semi-supervised learning

Security and Safety in Artificial Intelligence

Open Access

Table 4.

Performance of co-training with BBAug (AP and mAP)

	Semi-supervised_0.5_BBAug					Semi-supervised_0.7_BBAug					Semi-supervised_0.9_BBAug
		Car	Cyclist	Pedestrian	mAP (%)		Car	Cyclist	Pedestrian	mAP (%)		Car	Cyclist	Pedestrian	mAP (%)
	Clean	89.79	83.17	77.50	83.49	clean	87.99	67.85	72.26	76.04	Clean	88.13	72.64	75.40	78.72
White-box	Dpatch	89.75	83.13	77.50	83.46	Dpatch	87.94	67.62	71.10	75.55	Dpatch	87.92	72.63	75.40	78.65
White-box	Contextual patch	89.53	83.17	77.44	83.38	Contextual patch	87.88	67.45	72.03	75.79	Contextual patch	88.09	72.42	75.22	78.58

Black-box	Noise_snow	68.53	38.20	40.37	49.03	Noise_snow	67.57	37.49	42.07	49.04	Noise_snow	59.27	37.04	33.40	43.24
	Noise_fog	88.54	68.11	67.23	74.62	Noise_fog	79.53	56.07	55.77	63.79	Noise_fog	79.40	61.21	64.60	68.40
	Noise_frost	87.02	71.20	65.78	74.66	Noise_frost	77.79	53.34	59.98	63.70	Noise_frost	77.39	61.60	60.44	66.48
	Noise_gaussian	70.35	47.05	48.87	55.42	Noise_gaussian	68.91	36.36	44.07	49.78	Noise_gaussian	60.92	35.97	44.50	47.13
	Noise_shot	88.13	70.01	65.54	74.56	Noise_shot	79.34	53.80	54.26	62.46	Noise_shot	78.98	56.27	57.74	64.33
	Noise_impulse	61.20	35.34	40.06	45.54	Noise_impulse	60.35	24.72	35.91	40.32	Noise_impulse	52.03	20.39	30.69	34.37

	Average of b-box	77.30	54.99	54.64	62.31	Average of b-box	72.25	43.63	48.68	54.85	Average of b-box	68.00	45.41	48.56	53.99

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.