Fast Monocular Visual-Inertial Initialization with an Improved Iterative Strategy

Table 3

Result of the pose estimation in the 11 sequences of the EuRoC dataset. The reported values are the median after 10 executions for each sequence. The italic types indicate the best results. Abbreviations: Trans: translation; Ori: orientation; Ave: average.

Sequence	VI-ORBSLAM		VINS-MONO		OURS		ORB-SLAM3 (mono)
	Trans. (m)	Ori. (^〇)	Trans. (m)	Ori. (^〇)	Trans. (m)	Ori. (^〇)	Trans. (m)	Ori. (^〇)
V1_01_easy	0.027	2.112	0.047	2.733	0.031	2.657	0.043	2.422
V1_02_medium	0.028	1.692	0.066	1.732	0.017	1.646	0.016	1.673
V1_03_difficult			0.180	4.598	0.021	2.512	0.025	2.632
V2_01_easy	0.053	1.326	0.058	1.525	0.013	1.069	0.021	1.126
V2_02_medium	0.041	2.580	0.090	2.613	0.030	2.394	0.015	1.180
V2_03_difficult	0.074	4.012	0.244	5.532	0.049	3.496	0.037	4.362
MH_01_easy	0.075	2.251	0.084	2.310	0.031	2.036	0.042	2.155
MH_02_easy	0.084	2.523	0.105	2.753	0.015	2.408	0.053	3.043
MH_03_medium	0.087	1.353	0.074	1.262	0.026	1.156	0.043	1.935
MH_04_difficult	0.217	1.306	0.122	1.246	0.110	1.115	0.099	1.106
MH_05_difficult	0.082	0.559	0.147	0.395	0.056	0.511	0.071	0.529
Ave.	0.077	1.971	0.111	2.427	0.036	1.909	0.042	2.015