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

Security and Safety in Artificial Intelligence

Open Access

Table 2.

Performance under patch attack in a black-box manner (AP and mAP)

Cpatch in a black-box manner							Dpatch in a black-box manner
	YOLOv2	YOLOv4	SSD	Faster-RCNN	MoCov1	MoCov2		YOLOv2	YOLOv4	SSD	Faster-RCNN	MoCov1	MoCov2
					+Faster-RCNN	+Faster-RCNN						+Faster-RCNN	+Faster-RCNN
Clean	51.13	60.31	56.83	82.91	84.40	83.25	Clean	51.13	60.31	56.83	83.20	84.40	83.25
YOLOv2	46.80	53.07	53.28	82.91	84.36	82.86	YOLOv2	42.98	56.60	56.29	83.21	84.36	82.88
YOLOv4	50.27	54.14	56.80	82.90	84.37	83.01	YOLOv4	50.27	59.69	56.78	83.21	84.38	83.01
SSD	50.30	59.26	46.48	82.92	84.37	83.03	SSD	50.31	59.17	41.00	83.20	83.36	83.02
Faster-RCNN	48.50	55.13	56.80	81.96	83.36	82.94	Faster-RCNN	50.44	57.78	56.84	77.36	83.32	82.96
MoCov1+Faster-RCNN	50.41	57.65	56.82	82.91	84.20	82.99	MoCov1+Faster-RCNN	50.44	57.59	56.74	83.20	84.32	82.90
MoCov2+Faster-RCNN	50.26	56.07	56.42	82.93	83.40	82.97	MoCov2+Faster-RCNN	50.30	57.23	56.79	83.20	83.39	83.00

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.