Table 6.

Accuracy of EndoVis articulated robotic surgical instrument training data for the tracked point.

	Seq 1L	Seq 1R	Seq 2	Seq 3	Seq 4	Whole
Original annotation
In-view (IV) and Out-of-view (OV) Frame Number
IV	1107	1107	1096	1118	1056	5484
OV	0	0	29	6	67	102
Total	1107	1107	1125	1124	1123	5586
Accuracy ($Thres=20$ px)
Acc. (%)	85.00	92.86	90.60	88.10	36.55	79.01
Dist. (px)	7.42	7.07	7.41	9.64	9.26	8.00
Accuracy ($Thres=30$ px)
Acc. (%)	99.37	96.93	96.35	95.80	82.67	94.33
Dist. (px)	9.76	7.80	8.36	10.71	18.07	10.67
High quality annotation
In-view (IV) and Out-of-view (OV) frame number
IV	1107	1107	1099	1105	1066	5484
OV	0	0	26	19	57	102
Total	1107	1107	1125	1124	1123	5586
Accuracy ($Thres=20$ px)
Acc. (%)	100.0	99.73	98.91	98.28	95.78	98.56
Dist. (px)	4.89	9.87	3.29	4.31	11.13	6.65
Accuracy ($Thres=30$ px)
Acc. (%)	100.0	100.0	99.36	99.46	99.72	99.71
Dist. (px)	4.89	9.90	3.38	4.56	11.57	6.83