M^3: Dense Matching Meets Multi-View Foundation Models for Monocular Gaussian Splatting SLAM

Kerui Ren; Guanghao Li; Changjian Jiang; Yingxiang Xu; Tao Lu; Linning Xu; Junting Dong; Jiangmiao Pang

cs.CVMar 17, 2026

M^3: Dense Matching Meets Multi-View Foundation Models for Monocular Gaussian Splatting SLAM

Authors: Kerui Ren, Guanghao Li, Changjian Jiang, Yingxiang Xu, Tao Lu, Linning Xu, Junting Dong, Jiangmiao Pang,

Organizations: Shanghai Jiao Tong University · Shanghai Artificial Intelligence Laboratory · Fudan University · Shanghai Innovation Institute ·

Abstract

Streaming reconstruction from uncalibrated monocular video remains challenging, as it requires both high-precision pose estimation and computationally efficient online refinement in dynamic environments. While coupling 3D foundation models with SLAM frameworks is a promising paradigm, a critical bottleneck persists: most multi-view foundation models estimate poses in a feed-forward manner, yielding pixel-level correspondences that lack the requisite precision for rigorous geometric optimization. To address this, we present M^3, which augments the Multi-view foundation model with a dedicated Matching head to facilitate fine-grained dense correspondences and integrates it into a robust Monocular Gaussian Splatting SLAM. M^3 further enhances tracking stability by incorporating dynamic area suppression and cross-inference intrinsic alignment. Extensive experiments on diverse indoor and outdoor benchmarks demonstrate state-of-the-art accuracy in both pose estimation and scene reconstruction. Notably, M^3 reduces ATE RMSE by 64.3% compared to VGGT-SLAM 2.0 and outperforms ARTDECO by 2.11 dB in PSNR on the ScanNet++ dataset.

Figures & tables

Figure 1: We demonstrate the execution of our M 3 pipeline on a challenging manor sequence. Our approach delivers robust, high-precision pose estimation while achieving high-fidelity scene reconstruction from monocular video sequences. Please refer to our project page for more demos: https://city-super.github.io/M3/

Figure 2: The M 3 Pipeline. The framework consists of joint tracking and global optimization for pose estimation and a mapper for scene reconstruction. For monocular sequences, Pi3X processes retrieved historical keyframes and new frames in one inference to facilitate factor graph construction and keyframe selection. Following the Neural Gaussian and LOD architecture of ARTDECO [ 19 ] , Gaussians are initialized via Laplacian norm and optimized jointly with camera poses.

Dataset	DROID-SLAM	MASt3R-SLAM	VGGT-SLAM	VGGT-SLAM 2.0	ARTDECO	Ours	Pi3X
ScanNet++	0.623	0.346	1.081	0.182	0.137	0.065	-
ScanNetV2	0.217	0.106	0.852	0.073	0.141	0.051	-
Waymo	7.535	16.610	13.166	1.295	2.410	0.773	0.827
KITTI	8.212	18.073	10.985	2.521	2.938	0.890	1.441

Table 1: Tracking comparisons in terms of ATE RMSE (m). We compare against state-of-the-art SLAM frameworks on both indoor and outdoor datasets. The best and second-best results are highlighted in bold and underline , respectively.

Indoor-dataset	ScanNet++			ScanNetV2			VR-NeRF			Efficiency
Method	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	#GS	Time
MonoGS	18.56	0.732	0.582	21.88	0.827	0.508	14.69	0.565	0.728	25.2k	10.2 min
S3PO-GS	22.60	0.812	0.457	24.05	0.850	0.439	22.10	0.766	0.528	29.9k	17.1 min
HI-SLAM2	23.89	0.811	0.342	23.75	0.820	0.295	28.25	0.868	0.277	152.1k	6.9 min
OnTheFly-NVS	22.00	0.808	0.336	23.38	0.861	0.348	28.23	0.888	0.257	210.5k	1.4 min
ARTDECO	26.71	0.864	0.246	26.03	0.902	0.278	29.15	0.883	0.231	936.7k	5.1 min

Table 2: Rendering comparisons across indoor and outdoor datasets. We report visual quality metrics, the number of Gaussians, and average training time. The best and second-best results are highlighted in bold and underline , respectively.

Figure 3: Qualitative comparisons of rendering against on-the-fly reconstruction baselines across diverse datasets. M 3 preserves high-fidelity rendering details in challenging environments, particularly in regions highlighted by white rectangles.

Method	Ours				Depth-Anything-3				AnySplat
Dataset	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	GPU $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	GPU $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	GPU $\downarrow$
ScanNet++	27.789	0.873	0.211	22.4G	17.973	0.666	0.441	105.5G	22.337	0.757	0.236	94.6G
Waymo	28.346	0.870	0.273	23.2G	21.760	0.701	0.338	82.0G	21.974	0.729	0.341	73.1G

Table 3: Comparisons against feed-forward Gaussian Splatting methods. We evaluate reconstruction quality and efficiency on fixed-length sequences. Our optimization-based approach consistently outperforms SOTA feed-forward models.

Figure 4: Qualitative comparisons of rendering against feed-forward Gaussian Splatting methods. M 3 consistently preserves fine-grained visual details.

Figure 5: Effect of the Motion Map in dynamic environments. By detecting dynamic regions, the Motion Map enables moving objects to be excluded from the static reconstruction, thereby improving structural consistency in dynamic scenes.

Metrics	Full	w/o descriptor	w/o intr. align	w/o global opt.	$\rm{pi3x}\rightarrow\rm{pi3}$	split front&back-end	$L$ =16
ATE $\downarrow$	0.065	0.094	0.236	0.130	0.068	0.098	0.077
PSNR $\uparrow$	28.82	27.73	26.14	28.49	28.77	28.41	28.57
SSIM $\uparrow$	0.892	0.874	0.851	0.886	0.892	0.886	0.888
LPIPS $\downarrow$	0.190	0.221	0.258	0.201	0.191	0.203	0.199

Table 4: Ablation studies on the ScanNet++ dataset evaluating the effect of key design choices and hyperparameters.

Method	Ours (Pi3X)				Depth-Anything-3				VGGT
Dataset	ATE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	ATE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$	ATE $\downarrow$	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$
ScanNet++	0.065	28.82	0.892	0.190	0.079	28.09	0.882	0.211	0.106	27.33	0.869	0.234

Table 5: Comparison of Multi-view Foundation Models within the M 3 Framework. We evaluate the performance of different backbones on the ScanNet++ dataset [ 48 ] . Pi3X [ 41 ] and Depth-Anything-3 [ 22 ] represent the state-of-the-art in geometric priors, with Pi3X achieving the highest accuracy due to its superior cross-inference consistency.

Figure 6: Qualitative comparisons of camera trajectories against on-the-fly reconstruction baselines across diverse datasets. Our visualized trajectories align most closely with the ground truth (GT), exhibiting no significant divergence or inconsistency.

Figure 7: Qualitative comparisons of rendering against on-the-fly reconstruction baselines in outdoor scenes. Our rendered outputs exhibit superior clarity and are devoid of visual artifacts.

Figure 8: Qualitative comparisons of rendering against on-the-fly reconstruction baselines in indoor scenes. Our rendered outputs exhibit superior clarity and are devoid of visual artifacts.

Method	00777c41d4	126d03d821	1cbb105c6a	4808c4a397	546292a9db	7543973e1a	0a7cc12c0e
DROID-SLAM	1.677	0.231	0.304	0.253	0.386	0.414	1.285
MASt3R-SLAM	0.584	0.170	0.232	0.338	0.144	0.132	0.687
VGGT-SLAM	1.715	1.765	0.651	1.594	0.904	1.185	1.688
VGGT-SLAM 2.0	0.202	0.178	0.185	0.273	0.037	0.333	0.207
ARTDECO	0.072	0.057	0.148	0.086	0.043	0.176	0.455
Ours	0.073	0.305	0.025	0.054	0.087	0.053	0.016

Table 6: ATE RMSE (m) on the ScanNet++ Dataset

Method	scene0081_00	scene0354_00	scene0658_00	scene0671_00	scene0441_00
DROID-SLAM	0.084	0.529	0.171	0.032	0.068
MASt3R-SLAM	0.130	0.150	0.094	0.084	0.107
VGGT-SLAM	0.447	1.131	0.553	1.191	0.422
VGGT-SLAM 2.0	0.097	0.063	0.080	0.036	0.047
ARTDECO	0.106	0.543	0.060	0.039	0.052
Ours	0.046	0.093	0.044	0.029	0.040

Table 7: ATE RMSE (m) on the ScanNetV2 Dataset

Method	100613	106762	132384	13476	152706	153495	158686	163453	405841
DROID-SLAM	1.490	2.117	3.178	1.163	2.560	12.146	11.144	24.172	9.848
MASt3R-SLAM	17.464	32.556	48.649	13.496	22.706	4.029	3.278	4.548	2.767
VGGT-SLAM	25.969	1.247	1.607	12.910	33.078	17.520	14.157	1.455	10.554
VGGT-SLAM 2.0	0.927	1.516	1.756	1.710	0.620	1.029	1.677	1.543	0.880
ARTDECO	1.828	3.682	3.175	1.139	1.103	3.018	2.606	3.763	1.380
Ours	0.453	1.621	0.872	0.356	0.639	0.974	0.204	1.369	0.471

Table 8: ATE RMSE (m) on the Waymo Dataset

Method	00	02	03	05	06	07	08	10
DROID-SLAM	18.534	3.510	5.335	8.316	10.982	6.045	8.555	4.417
MASt3R-SLAM	26.038	56.199	1.519	21.461	1.621	1.943	31.022	4.783
VGGT-SLAM	2.205	35.429	2.137	1.824	20.523	22.128	0.543	3.093
VGGT-SLAM 2.0	2.300	3.751	2.213	1.811	3.374	3.212	0.557	2.952
ARTDECO	1.913	3.144	1.790	2.552	8.136	1.745	1.962	2.259
Ours	0.376	1.284	0.944	1.062	1.395	0.511	0.465	1.083

Table 9: ATE RMSE (m) on the KITTI Dataset

Method	00777c41d4	126d03d821	1cbb105c6a	4808c4a397	546292a9db	7543973e1a	0a7cc12c0e
MonoGS	11.57	21.14	21.64	21.98	11.20	21.37	18.51
S3PO-GS	17.78	23.38	25.23	22.64	20.13	24.21	20.67
HI-SLAM2	18.45	25.51	23.09	23.95	22.77	26.79	23.35
OnTheFly-NVS	19.49	19.43	19.52	18.57	26.13	20.86	21.55
ARTDECO	25.33	28.64	26.93	30.17	25.53	27.69	28.88
Ours	25.20	29.56	28.29	30.12	23.43	30.67	31.14

Table 10: PSNR on the ScanNet++ Dataset

Method	00777c41d4	126d03d821	1cbb105c6a	4808c4a397	546292a9db	7543973e1a	0a7cc12c0e
MonoGS	0.474	0.792	0.800	0.852	0.367	0.823	0.813
S3PO-GS	0.759	0.817	0.847	0.854	0.671	0.853	0.818
HI-SLAM2	0.608	0.870	0.804	0.869	0.747	0.893	0.831
OnTheFly-NVS	0.655	0.779	0.781	0.823	0.847	0.812	0.835
ARTDECO	0.807	0.904	0.874	0.915	0.776	0.873	0.848
Ours	0.807	0.907	0.904	0.931	0.776	0.927	0.929

Table 11: SSIM on the ScanNet++ Dataset

Method	00777c41d4	126d03d821	1cbb105c6a	4808c4a397	546292a9db	7543973e1a	0a7cc12c0e
MonoGS	0.828	0.458	0.476	0.396	0.733	0.490	0.491
S3PO-GS	0.532	0.382	0.388	0.371	0.547	0.402	0.437
HI-SLAM2	0.582	0.194	0.331	0.292	0.370	0.301	0.329
OnTheFly-NVS	0.419	0.365	0.376	0.330	0.214	0.343	0.302
ARTDECO	0.256	0.185	0.237	0.175	0.309	0.254	0.277
Ours	0.244	0.176	0.183	0.150	0.281	0.174	0.147

Table 12: LPIPS on the ScanNet++ Dataset

Method	scene0081_00	scene0354_00	scene0658_00	scene0671_00	scene0441_00
MonoGS	23.06	19.15	21.68	22.25	21.86
S3PO-GS	27.58	20.47	22.63	23.01	22.13
HI-SLAM2	28.03	23.41	20.79	25.51	21.81
OnTheFly-NVS	28.58	21.75	19.27	26.63	18.96
ARTDECO	33.24	23.14	20.55	28.64	19.65
Ours	33.44	30.87	20.18	29.73	19.33

Table 13: PSNR on the ScanNetV2 Dataset

Method	scene0081_00	scene0354_00	scene0658_00	scene0671_00	scene0441_00
MonoGS	0.831	0.837	0.838	0.827	0.860
S3PO-GS	0.861	0.848	0.853	0.838	0.865
HI-SLAM2	0.845	0.872	0.807	0.837	0.845
OnTheFly-NVS	0.871	0.868	0.858	0.877	0.881
ARTDECO	0.901	0.890	0.903	0.906	0.904
Ours	0.896	0.941	0.895	0.905	0.902

Table 14: SSIM on the ScanNetV2 Dataset

Method	scene0081_00	scene0354_00	scene0658_00	scene0671_00	scene0441_00
MonoGS	0.513	0.449	0.519	0.484	0.529
S3PO-GS	0.429	0.399	0.457	0.443	0.499
HI-SLAM2	0.199	0.266	0.340	0.230	0.289
OnTheFly-NVS	0.350	0.309	0.381	0.311	0.346
ARTDECO	0.286	0.278	0.306	0.262	0.300
Ours	0.304	0.174	0.320	0.262	0.287

Table 15: LPIPS on the ScanNetV2 Dataset

Method	CBD_Building_01	HKU_Campus	Red_Sculpture	Retail_Street	SYSU
MonoGS	19.38	20.09	15.40	18.34	18.63
S3PO-GS	20.83	22.95	18.13	21.74	19.48
HI-SLAM2	21.75	22.29	17.36	27.11	23.49
OnTheFly-NVS	18.49	22.71	17.60	17.66	19.66
ARTDECO	26.55	25.64	21.09	26.90	23.35
Ours	25.09	26.84	22.75	25.98	26.72

Table 16: PSNR on the FAST-LIVO2 Dataset

Method	CBD_Building_01	HKU_Campus	Red_Sculpture	Retail_Street	SYSU
MonoGS	0.683	0.586	0.520	0.562	0.595
S3PO-GS	0.720	0.634	0.571	0.683	0.618
HI-SLAM2	0.794	0.662	0.594	0.897	0.690
OnTheFly-NVS	0.685	0.649	0.589	0.552	0.629
ARTDECO	0.861	0.750	0.702	0.848	0.734
Ours	0.818	0.786	0.756	0.826	0.810

Table 17: SSIM on the FAST-LIVO2 Dataset

Method	CBD_Building_01	HKU_Campus	Red_Sculpture	Retail_Street	SYSU
MonoGS	0.645	0.759	0.770	0.720	0.698
S3PO-GS	0.413	0.550	0.551	0.410	0.571
HI-SLAM2	0.275	0.493	0.573	0.084	0.346
OnTheFly-NVS	0.470	0.427	0.461	0.517	0.476
ARTDECO	0.246	0.339	0.364	0.200	0.319
Ours	0.347	0.302	0.321	0.214	0.281

Table 18: LPIPS on the FAST-LIVO2 Dataset

Method	00	02	03	05	06	07	08	10
MonoGS	15.06	13.37	17.80	15.98	15.54	11.21	13.04	14.78
S3PO-GS	18.55	18.55	19.93	19.12	17.24	18.99	18.65	16.57
HI-SLAM2	18.97	23.58	23.35	21.09	18.64	21.06	19.80	21.24
OnTheFly-NVS	15.11	16.65	18.72	18.33	16.77	17.56	18.57	15.01
ARTDECO	21.29	22.69	22.16	22.30	21.39	21.14	20.91	19.86
Ours	22.82	22.29	23.26	23.59	23.48	22.35	21.86	20.06

Table 19: PSNR on the KITTI Dataset

Method	00	02	03	05	06	07	08	10
MonoGS	0.537	0.435	0.496	0.494	0.553	0.436	0.472	0.499
S3PO-GS	0.669	0.668	0.560	0.605	0.562	0.669	0.618	0.523
HI-SLAM2	0.659	0.750	0.719	0.659	0.650	0.723	0.681	0.681
OnTheFly-NVS	0.576	0.523	0.563	0.620	0.564	0.641	0.675	0.498
ARTDECO	0.759	0.740	0.660	0.731	0.685	0.745	0.709	0.642
Ours	0.803	0.681	0.676	0.755	0.726	0.764	0.745	0.666

Table 20: SSIM on the KITTI Dataset

Method	00	02	03	05	06	07	08	10
MonoGS	0.719	0.781	0.714	0.743	0.731	0.796	0.832	0.770
S3PO-GS	0.323	0.326	0.487	0.414	0.491	0.383	0.376	0.530
HI-SLAM2	0.359	0.248	0.301	0.376	0.442	0.270	0.310	0.337
OnTheFly-NVS	0.458	0.480	0.431	0.411	0.479	0.371	0.333	0.466
ARTDECO	0.292	0.263	0.341	0.247	0.381	0.252	0.333	0.299
Ours	0.223	0.321	0.303	0.259	0.314	0.280	0.285	0.318

Table 21: LPIPS on the KITTI Dataset

Method	100613	106762	132384	13476	152706	153495	158686	163453	405841
MonoGS	20.07	21.32	23.52	19.45	21.62	13.85	19.66	19.46	16.46
S3PO-GS	23.86	27.16	26.47	23.82	26.08	25.65	24.15	23.05	25.53
HI-SLAM2	25.79	27.35	24.76	25.00	25.63	24.33	22.42	18.65	18.60
OnTheFly-NVS	29.84	31.23	26.79	25.36	27.48	29.39	24.68	23.71	23.20
ARTDECO	29.16	31.92	31.10	26.54	31.57	27.38	26.93	24.43	28.62
Ours	30.21	27.49	30.35	28.91	30.97	27.55	28.63	26.62	29.75

Table 22: PSNR on the Waymo Dataset

Method	100613	106762	132384	13476	152706	153495	158686	163453	405841
MonoGS	0.745	0.808	0.851	0.698	0.786	0.661	0.700	0.739	0.690
S3PO-GS	0.794	0.852	0.874	0.749	0.816	0.818	0.768	0.779	0.827
HI-SLAM2	0.811	0.880	0.881	0.742	0.791	0.797	0.743	0.615	0.624
OnTheFly-NVS	0.902	0.915	0.893	0.812	0.864	0.902	0.803	0.803	0.784
ARTDECO	0.890	0.923	0.920	0.779	0.896	0.862	0.835	0.810	0.870
Ours	0.888	0.886	0.915	0.863	0.885	0.867	0.864	0.865	0.889

Table 23: SSIM on the Waymo Dataset

Method	100613	106762	132384	13476	152706	153495	158686	163453	405841
MonoGS	0.623	0.539	0.456	0.754	0.667	0.728	0.648	0.670	0.638
S3PO-GS	0.445	0.368	0.327	0.536	0.505	0.479	0.429	0.449	0.392
HI-SLAM2	0.208	0.162	0.157	0.319	0.307	0.298	0.295	0.403	0.435
OnTheFly-NVS	0.243	0.248	0.301	0.304	0.294	0.259	0.333	0.332	0.384
ARTDECO	0.234	0.203	0.221	0.334	0.216	0.275	0.254	0.283	0.250
Ours	0.266	0.305	0.262	0.264	0.295	0.306	0.256	0.290	0.245

Table 24: LPIPS on the Waymo Dataset

Method	appartment-26-2	kitchen-26-1	kitchen-26-2	kitchen-26-3
MonoGS	18.30	15.93	10.70	17.41
S3PO-GS	25.16	24.58	24.91	20.70
HI-SLAM2	27.65	29.27	28.79	26.71
OnTheFly-NVS	30.87	32.23	28.79	25.59
ARTDECO	31.98	31.51	31.37	28.86
Ours	32.26	28.60	29.94	28.29

Table 25: PSNR on the VRNeRF Dataset

Method	appartment-26-2	kitchen-26-1	kitchen-26-2	kitchen-26-3
MonoGS	0.678	0.539	0.509	0.527
S3PO-GS	0.834	0.807	0.834	0.680
HI-SLAM2	0.866	0.881	0.908	0.852
OnTheFly-NVS	0.912	0.922	0.921	0.857
ARTDECO	0.915	0.912	0.937	0.895
Ours	0.916	0.879	0.915	0.871

Table 26: SSIM on the VRNeRF Dataset

Method	appartment-26-2	kitchen-26-1	kitchen-26-2	kitchen-26-3
MonoGS	0.625	0.711	0.744	0.658
S3PO-GS	0.441	0.468	0.414	0.557
HI-SLAM2	0.350	0.269	0.220	0.309
OnTheFly-NVS	0.307	0.225	0.223	0.246
ARTDECO	0.232	0.205	0.184	0.200
Ours	0.192	0.270	0.189	0.206

Table 27: LPIPS on the VRNeRF Dataset

Visual SLAM

Visual simultaneous localization and mapping (SLAM) estimates camera motion while building maps from images for navigation and scene understanding.

See all 103 papers

GaussianFlow SLAM: Monocular Gaussian Splatting SLAM Guided by GaussianFlow
Apr 17, 2026Dong-Uk Seo, Jinwoo Jeon, Eungchang Mason Lee +13D Gaussian Splatting 3DGS SLAM
School of Electrical Engineering, KAIST (Korea Advanced Institute of Science and Technology), Daejeon 34141, Republic of Korea
AIM-SLAM: Dense Monocular SLAM via Adaptive and Informative Multi-View Keyframe Prioritization with Foundation Model
Mar 5, 2026Jinwoo Jeon, Dong-Uk Seo, Eungchang Mason Lee +1Simultaneous Localization and Mapping Visual SLAM
The School of Electrical Engineering, KAIST (Korea Advanced Institute of Science and Technology), Daejeon, 34141, Republic of Korea · KAIST InnoCORE LLM, KAIST, Daejeon, 34141, Republic of Korea
Flash-Mono: Feed-Forward Accelerated Gaussian Splatting Monocular SLAM
Apr 3, 2026Zicheng Zhang, Ke Wu, Xiangting Meng +33DGS SLAM 3D Gaussian Splatting
Fudan University · ShanghaiTech University · Shanghai Jiao Tong University

Apr 17, 2026J/K move · Enter open · S save

GaussianFlow SLAM: Monocular Gaussian Splatting SLAM Guided by GaussianFlow

Dong-Uk Seo, Jinwoo Jeon, Eungchang Mason Lee, Hyun Myung

School of Electrical Engineering, KAIST (Korea Advanced Institute of Science and Technology), Daejeon 34141, Republic of Korea

3D Gaussian Splatting 3DGS SLAM Camera Pose Estimation Visual SLAM Gaussian Splatting

Open paper AskPDF arXiv

Gaussian splatting has recently gained traction as a compelling map representation for SLAM systems, enabling dense and photo-realistic scene modeling. However, its application to monocular SLAM remains challenging due to the lack of reliable geometric cues from monocular input. Without geometric supervision, mapping or tracking could fall in local-minima, resulting in structural degeneracies and inaccuracies. To address this challenge, we propose GaussianFlow SLAM, a monocular 3DGS-SLAM that leverages optical flow as a geometry-aware cue to guide the optimization of both the scene structure and camera poses. By encouraging the projected motion of Gaussians, termed GaussianFlow, to align with the optical flow, our method introduces consistent structural cues to regularize both map reconstruction and pose estimation. Furthermore, we introduce normalized error-based densification and pruning modules to refine inactive and unstable Gaussians, thereby contributing to improved map quality and pose accuracy. Experiments conducted on public datasets demonstrate that our method achieves superior rendering quality and tracking accuracy compared with state-of-the-art algorithms. The source code is available at: https://github.com/url-kaist/gaussianflow-slam.

Abstract

Visual SLAM

Explore similar work

GaussianFlow SLAM: Monocular Gaussian Splatting SLAM Guided by GaussianFlow

AIM-SLAM: Dense Monocular SLAM via Adaptive and Informative Multi-View Keyframe Prioritization with Foundation Model

Flash-Mono: Feed-Forward Accelerated Gaussian Splatting Monocular SLAM