[{"data":1,"prerenderedAt":102},["ShallowReactive",2],{"method-hislam2_2025":3},{"method":4,"reference":64,"equipment":88,"figures":101,"results":94},{"id":5,"label":6,"shortName":7,"title":8,"year":9,"era":10,"cluster":11,"scope":12,"keyIdeaZh":13,"keyIdeaEn":14,"fulltextStatus":15,"publicationStatus":16,"recommendation":17,"constructionRelevance":18,"validationEnvironment":19,"strengths":23,"limitations":28,"sensors":33,"platform":35,"estimator":40,"association":41,"timeModel":42,"deskew":43,"loopClosure":44,"globalOptimization":45,"mapRepresentation":46,"prior":47,"outputGeometry":48,"compute":49,"codeUrl":50,"codeLicense":51,"relatedVersions":52},"hislam2_2025","Zhang et al., 2025","HI-SLAM2","HI-SLAM2: Geometry-Aware Gaussian SLAM for Fast Monocular Scene Reconstruction",2025,"recent","C09","full_slam_with_global_correction","HI-SLAM2 是只用單眼 RGB 的三維高斯 SLAM：追蹤端沿用 DROID-SLAM 的學習式光流與稠密光束法平差，並以每張影像 2x2 的尺度網格把 Omnidata 單眼深度先驗對齊到估計深度，以修正先驗中隨位置變化的尺度失真。偵測到迴圈時以 Sim(3) 位姿圖平差同時修正位姿與尺度漂移，並依錨定關鍵影格的更新直接變形高斯，使地圖即時保持一致；離線階段再做完整光束法平差與位姿、高斯聯合最佳化。地圖以高斯表示並以射線與高斯交點計算無偏深度，最後由渲染深度經 TSDF 融合得到網格。","Monocular RGB Gaussian-splatting SLAM that couples DROID-SLAM tracking with grid-based scale alignment of monocular depth priors, Sim(3) pose-graph loop closure that deforms keyframe-anchored Gaussians online, and offline full BA plus joint pose-map refinement; meshes via TSDF fusion of rendered depth.","full_text_reviewed","peer_reviewed_published","main_body","論文在大型工廠廠房以機器人左相機拍攝 4073 影格進行實測，並以文獻 [81] 的公分級攝影測量參考比對軌跡（僅圖 14 定性呈現），也在具雷射掃描真值的 ScanNet++ iPhone 序列上展示重建；作者來自斯圖加特大學攝影測量研究所。其前身 HI-SLAM 是語料中 AEC 領域 SLAM 回顧 [li2026slamgenerationaec] 唯一提及的神經 SLAM。定量幾何精度仍只在合成 Replica 上報告（精度約 1.6 cm），在施工現場或以全測站、TLS 驗證的量測尚未出現；作者也指出城市尺度與動態環境仍是限制。",[20,21,22],"public_benchmark","simulation","controlled_experiment",[24,25,26,27],"Best Replica geometry among RGB-only methods: accuracy 1.57 cm, completeness 3.49 cm, completion ratio 85.25% (Table IV)","Lowest average ATE on Replica (0.26 cm), ScanNet (7.07 cm) and Waymo Open (0.457 m) among compared methods (Tables I-III)","Online operation at 22 FPS on Replica with 12 s offline refinement (Sec. IV-H)","Factory-hall robot sequence processed in about half the runtime of DROID-SLAM + 3DGS with better geometry; trajectory compared against a photogrammetric reference (Sec. IV-I, Fig. 14)",[29,30,31,32],"Proximity-based loop detection is not robust to view occlusions and textureless regions (ETH3D); learned place recognition is suggested (Sec. V)","Mapping quality can degrade in city-scale scenes because of the limited optimization budget; submaps suggested (Sec. V)","Assumes static environments (Sec. V)","On ETH3D, 6 of 61 sequences failed in complete darkness and 4 more because of lighting changes and occlusions (Sec. IV-D)",[34],"monocular RGB camera",[36,37,38,39],"simulation (Replica)","vehicle (Waymo Open front camera)","ground robot with stereo cameras, left camera only (self-collected factory hall; robot type not described)","handheld (ScanNet++ iPhone sequences)","DROID-SLAM-based recurrent optical flow and dense bundle adjustment (damped Gauss-Newton) on a keyframe graph, interleaved with joint depth and scale alignment (JDSA) that fits a 2x2 scale grid to each monocular depth prior; Sim(3) pose-graph BA on loop closure; offline full BA and joint pose and 3DGS refinement with Adam (Sec. III-B to III-E)","dense learned optical-flow correspondences with confidence weights between co-visible keyframes (Sec. III-B)","discrete poses (keyframes)","not_applicable","proximity-based: candidates with optical-flow distance below a threshold, orientation difference below a threshold and index gap beyond the local window; closed by online Sim(3) pose-graph bundle adjustment that also corrects scale drift (Sec. III-C)","online Sim(3) pose-graph BA with relative-pose factors from retained dense correspondences, offline full BA over all overlapping keyframe pairs, then joint optimization of Gaussians, poses and exposure (Sec. III-C, III-E)","3D Gaussian splatting with RGB colours (no spherical harmonics), unbiased ray-Gaussian intersection depth, Gaussians anchored to keyframes and deformed with their Sim(3) updates; random downsampling (factor 32), densification and pruning; map grows without a predefined scene bound (Sec. III-D)","pretrained DROID-SLAM flow network and Omnidata monocular depth and normal priors (Sec. IV-A)","3DGS map and a mesh from TSDF fusion of rendered depth maps; rendered colour and depth (Sec. III)","NVIDIA RTX 4090 and Intel Core i9-12900K; online tracking, loop closing and mapping at 22 FPS on Replica and above 10 FPS on ScanNet; offline refinement 12 s on Replica and a few minutes on ScanNet (Sec. IV-A, IV-H)","https:\u002F\u002Fgithub.com\u002FWillyzw\u002FHI-SLAM2","BSD-3-Clause (LICENSE file checked; copyright line inherited from Princeton Vision & Learning Lab)",[53,57,61],{"relation":54,"title":55,"doi_or_url":56},"preprint","arXiv:2411.17982v3","https:\u002F\u002Farxiv.org\u002Fabs\u002F2411.17982",{"relation":58,"title":59,"doi_or_url":60},"predecessor","HI-SLAM (IEEE RA-L 2024)","https:\u002F\u002Fdoi.org\u002F10.1109\u002FLRA.2023.3347131",{"relation":62,"title":63,"doi_or_url":50},"code_release","Willyzw\u002FHI-SLAM2",{"id":5,"kind":65,"shortName":7,"title":8,"authors":66,"year":9,"venue":73,"venueType":74,"publisher":75,"volumeIssuePages":76,"doi":77,"arxivId":78,"url":79,"firstPublicDate":80,"publicationStatus":16,"metadataStatus":81,"fulltextStatus":15,"era":10,"classicReason":43,"codeUrl":50,"cluster":11,"topics":82,"mdpi":83,"verification":84,"label":6,"fulltextRoute":85,"versionRead":86,"addedByCensus":87},"method",[67,68,69,70,71,72],"Wei Zhang","Qing Cheng","David Skuddis","Niclas Zeller","Daniel Cremers","Norbert Haala","IEEE Transactions on Robotics","journal","IEEE","41:6478-6493","10.1109\u002Ftro.2025.3626627","2411.17982","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002FTRO.2025.3626627","2024-11-27","metadata_verified",[11],false,"confirmed","arXiv","arXiv v3 (2411.17982v3, 2 Feb 2026), posted after acceptance with the T-RO journal reference; not compared line by line with the IEEE version of record",true,[89,97],{"category":90,"model":91,"canonical":92,"role":93,"dataset":94,"specs":95,"locator":96},"compute","Nvidia RTX 4090","NVIDIA RTX 4090","compute for runtime",null,"GPU for all evaluations","Sec. IV-A",{"category":90,"model":98,"canonical":99,"role":93,"dataset":94,"specs":100,"locator":96},"Intel Core i9-12900K","Intel Core i9 12900K","CPU for all evaluations",[],1790510658261]