[{"data":1,"prerenderedAt":559},["ShallowReactive",2],{"method-loopyslam2024":3},{"method":4,"reference":60,"equipment":82,"figures":112,"results":113},{"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":22,"limitations":27,"sensors":36,"platform":38,"estimator":39,"association":40,"timeModel":41,"deskew":42,"loopClosure":43,"globalOptimization":44,"mapRepresentation":45,"prior":46,"outputGeometry":47,"compute":48,"codeUrl":49,"codeLicense":50,"relatedVersions":51},"loopyslam2024","Liso et al., 2024","Loopy-SLAM","Loopy-SLAM: Dense Neural SLAM with Loop Closures",2024,"recent","C09","full_slam_with_global_correction","Loopy-SLAM 在 Point-SLAM 的神經點雲上加入子地圖、詞袋式全域地點辨識與穩健位姿圖最佳化，迴圈閉合後直接剛性平移子地圖中的點以修正地圖，毋須保存全部歷史影格。作者未研究光束調整精修，且實作尚非即時。","Adds submaps, BoW place recognition and robust pose-graph optimization to a neural point-cloud SLAM so maps can be corrected after loop closure.","full_text_reviewed","peer_reviewed_published","background","論文未涉及營建場域；資料為 Replica、TUM-RGBD、ScanNet。",[20,21],"simulation","public_benchmark",[23,24,25,26],"Replica: lowest average ATE RMSE (0.29 cm vs 0.35 cm GO-SLAM and 0.52 cm Point-SLAM) and best average depth L1 (0.35 cm) and F1 at 1 cm (90.77%) (Table 1, Table 11)","TUM-RGBD: average ATE 3.85 cm vs 8.92 cm for Point-SLAM with about 14% more scene points (Table 2, Table 6)","ScanNet: lowest ATE on the multi-room scene 54 (7.5 cm) and drift-reduced meshes compared with Point-SLAM and ESLAM (Table 3, Fig. 4)","Map correction without storing all input frames (abstract; Sec. 3)",[28,29,30,31,32,33,34,35],"Not real-time; Python implementation (Limitations)","Effect of bundle adjustment not studied (App. D)","Place recognition could be improved with learned variants (Limitations)","No relocalization (Limitations)","Frame-to-model tracking alone is not the most robust choice; the authors suggest combining it with frame-to-frame tracking (Limitations)","Behind GO-SLAM on the ScanNet average (7.7 vs 7.0 cm Avg.-9) and behind ORB-SLAM2 (1.98 cm average) on TUM-RGBD (Table 2, Table 3)","Global registration takes about 12 s per registration with the default stopping criteria (Sec. 4.4, App. E Table 9)","Evaluation caveats: ScanNet reference poses come from BundleFusion, and meshes are ICP-aligned to ground truth before precision and recall (Sec. 4 Datasets, App. C)",[37],"RGB-D",[],"frame-to-model tracking on neural point submaps + robust pose graph optimization (line process) over global keyframes","Frame-to-model tracking by minimizing depth and colour re-rendering losses on the active submap (Point-SLAM style); loop edges from coarse-to-fine dense registration of submap surfaces: all depth frames of a submap are TSDF-fused and points sampled on the marching-cubes surface, FPFH features with RANSAC give the coarse alignment and ICP on full-resolution clouds refines it; loop edges pre-filtered by constraint translation magnitude and a fitness (overlap) score (Sec. 3.1, 3.2)","discrete poses","not_applicable","Global place recognition with a bag-of-visual-words database (DBoW3) queried when a submap is completed (top K = 4 on Replica, 1 on TUM-RGBD and ScanNet, dynamic similarity threshold); on ScanNet a PGO is triggered on average every 151 frames (Sec. 3.2, Sec. 4, App. E Table 8)","Robust pose graph optimization of global-keyframe corrections with a dense surface-registration objective and a line process weighting loop edges, solved with Levenberg-Marquardt in two stages; submaps and frame poses then corrected rigidly by shifting points; feature fusion and colour and geometry feature refinement at the end of capture; bundle adjustment not studied (Sec. 3.2, App. D)","submaps of neural point clouds","none","Mesh from TSDF fusion (1 cm voxels) of depth and colour rendered every fifth frame along the estimated trajectory, following Point-SLAM; rendered RGB-D images (Sec. 4 implementation details)","PyTorch and Open3D through Python bindings, not optimized for real time; all experiments on NVIDIA GPUs with at most 12 GB memory; Replica office 0: tracking 0.85 s and mapping 9.85 s per frame (identical to the Point-SLAM row); on fr1 desk 7 PGOs of about 1 ms each, each needing about 8 global registrations of about 12 s on an Intel Core i7-11800H (Sec. 4.4, Table 5, App. B, App. E)","https:\u002F\u002Fgithub.com\u002Feriksandstroem\u002FLoopy-SLAM","Apache-2.0",[52,56],{"relation":53,"title":54,"doi_or_url":55},"preprint","arXiv:2402.09944","https:\u002F\u002Farxiv.org\u002Fabs\u002F2402.09944",{"relation":57,"title":58,"doi_or_url":59},"project_page","Loopy-SLAM project page repository (no source code)","https:\u002F\u002Fgithub.com\u002Fnotchla\u002FLoopy-SLAM",{"id":5,"kind":61,"shortName":7,"title":8,"authors":62,"year":9,"venue":68,"venueType":69,"publisher":70,"volumeIssuePages":71,"doi":72,"arxivId":73,"url":74,"firstPublicDate":75,"publicationStatus":16,"metadataStatus":76,"fulltextStatus":15,"era":10,"classicReason":42,"codeUrl":49,"cluster":11,"topics":77,"mdpi":78,"verification":79,"label":6,"fulltextRoute":80,"versionRead":81,"addedByCensus":78},"method",[63,64,65,66,67],"Lorenzo Liso","Erik Sandström","Vladimir Yugay","Luc Van Gool","Martin R. Oswald","2024 IEEE\u002FCVF Conference on Computer Vision and Pattern Recognition (CVPR)","conference","IEEE","pp. 20363-20373","10.1109\u002Fcvpr52733.2024.01925","2402.09944","https:\u002F\u002Fapi.crossref.org\u002Fworks\u002F10.1109\u002Fcvpr52733.2024.01925","2024-02-14","metadata_verified",[11],false,"corrected","arXiv","arXiv v2 (2024-06-10) including supplementary material; main-paper key values (3.85 cm TUM average, 0.29 cm Replica ATE, 90.8 F1, 12 GB, no relocalization) cross-checked against the CVF open-access accepted version",[83,90,95,99,105],{"category":84,"model":85,"canonical":85,"role":86,"dataset":87,"specs":88,"locator":89},"compute","Nvidia GPUs with a maximum memory of 12 GB (models not named)","compute for runtime",null,"all results gathered on various GPUs of at most 12 GB","Sec. 4.4; App. B",{"category":84,"model":91,"canonical":91,"role":86,"dataset":92,"specs":93,"locator":94},"11th Gen Intel Core i7-11800H","TUM-RGBD","CPU used for the 12 s per registration timing","App. E",{"category":84,"model":96,"canonical":96,"role":86,"dataset":92,"specs":97,"locator":98},"AMD EPYC 7742","processor used for the global-registration ablation timings","App. E, Table 9",{"category":100,"model":101,"canonical":101,"role":102,"dataset":92,"specs":103,"locator":104},"other","external motion capture system (not named)","reference or ground truth","source of TUM-RGBD ground-truth poses","Sec. 4 Datasets",{"category":106,"model":107,"canonical":107,"role":108,"dataset":109,"specs":110,"locator":111},"rgbd","RGBD camera (model not named)","method input","Replica; TUM-RGBD; ScanNet","RGBD stream is the only input; TUM-RGBD and ScanNet are real-world data, the Replica trajectories come from a simulated RGBD sensor","Abstract; Fig. 2; Sec. 4 Datasets",[],{"totalRows":114,"groupCount":115,"groups":116,"others":548},19,6,[117,339,427,490],{"slug":118,"group":119,"sourceId":5,"sourceLabel":6,"table":120,"selfRows":115,"metrics":121,"seqs":127,"entrants":141,"cells":184,"outcomes":333,"locators":334,"hardware":335,"wordings":336,"notes":337},"loopyslam2024-table-2","loopyslam2024:Table 2","Table 2",[122],{"label":123,"unit":124,"statistic":125,"alignment":126},"ATE RMSE","cm","RMSE","not_reported",[128,131,133,135,137,139],{"dataset":92,"sequence":129,"environment":130},"fr1\u002Fdesk","TUM-RGBD real-world RGBD sequences (fr1\u002Fdesk, fr1\u002Fdesk2, fr1\u002Froom, fr2\u002Fxyz, fr3\u002Foffice); ground truth from an external motion capture system",{"dataset":92,"sequence":132,"environment":130},"fr2\u002Fxyz",{"dataset":92,"sequence":134,"environment":130},"fr3\u002Foffice",{"dataset":92,"sequence":136,"environment":130},"fr1\u002Fdesk2",{"dataset":92,"sequence":138,"environment":130},"fr1\u002Froom",{"dataset":92,"sequence":140,"environment":130},"Avg.",[142,144,148,150,152,155,158,161,163,165,168,171,174,177,180,182],{"name":143,"methodId":87,"linkable":78,"proposed":78,"self":78},"DI-Fusion [21]",{"name":145,"methodId":146,"linkable":147,"proposed":78,"self":78},"NICE-SLAM [77]","niceslam2022",true,{"name":149,"methodId":87,"linkable":78,"proposed":78,"self":78},"Vox-Fusion [73]",{"name":151,"methodId":87,"linkable":78,"proposed":78,"self":78},"MIPS-Fusion [57] (LC)",{"name":153,"methodId":154,"linkable":147,"proposed":78,"self":78},"Point-SLAM [45]","pointslam2023",{"name":156,"methodId":157,"linkable":147,"proposed":78,"self":78},"ESLAM [28]","eslam2023",{"name":159,"methodId":160,"linkable":147,"proposed":78,"self":78},"Co-SLAM [61]","coslam2023",{"name":162,"methodId":87,"linkable":78,"proposed":78,"self":78},"GO-SLAM [76] (LC)",{"name":164,"methodId":5,"linkable":147,"proposed":147,"self":147},"Loopy-SLAM (Ours, LC)",{"name":166,"methodId":167,"linkable":147,"proposed":78,"self":78},"BAD-SLAM [48] (LC)","badslam2019",{"name":169,"methodId":170,"linkable":147,"proposed":78,"self":78},"Kintinuous [69] (LC)","kintinuous2015",{"name":172,"methodId":173,"linkable":147,"proposed":78,"self":78},"ORB-SLAM2 [34] (LC)","orbslam2_2017",{"name":175,"methodId":176,"linkable":147,"proposed":78,"self":78},"ElasticFusion [71] (LC)","elasticfusion2015",{"name":178,"methodId":179,"linkable":147,"proposed":78,"self":78},"BundleFusion [13] (LC)","bundlefusion2017",{"name":181,"methodId":87,"linkable":78,"proposed":78,"self":78},"Cao et al. [7] (LC)",{"name":183,"methodId":87,"linkable":78,"proposed":78,"self":78},"Yan et al. [72] (LC)",[185,189,192,194,196,199,202,204,206,209,211,212,214,216,218,220,221,223,225,227,229,231,233,235,237,239,241,243,245,247,249,251,253,254,257,259,261,263,266,268,270,272,274,276,278,280,281,284,286,288,290,291,293,296,298,300,302,303,305,308,310,312,314,316,318,320,321,322,324,325,327,329,331],[186,186,186,187,188,186,188,188,186],0,4.4,-1,[186,186,190,191,188,186,188,188,186],1,2,[186,186,191,193,188,186,188,188,186],5.8,[190,186,186,195,188,186,188,188,186],4.26,[190,186,197,198,188,186,188,188,186],3,4.99,[190,186,200,201,188,186,188,188,186],4,34.49,[190,186,190,203,188,186,188,188,186],6.19,[190,186,191,205,188,186,188,188,186],3.87,[190,186,207,208,188,186,188,188,186],5,10.76,[191,186,186,210,188,186,188,188,186],3.52,[191,186,197,115,188,186,188,188,186],[191,186,200,213,188,186,188,188,186],19.53,[191,186,190,215,188,186,188,188,186],1.49,[191,186,191,217,188,186,188,188,186],26.01,[191,186,207,219,188,186,188,188,186],11.31,[197,186,186,197,188,186,188,188,186],[197,186,190,222,188,186,188,188,186],1.4,[197,186,191,224,188,186,188,188,186],4.6,[200,186,186,226,188,186,188,188,186],4.34,[200,186,197,228,188,186,188,188,186],4.54,[200,186,200,230,188,186,188,188,186],30.92,[200,186,190,232,188,186,188,188,186],1.31,[200,186,191,234,188,186,188,188,186],3.48,[200,186,207,236,188,186,188,188,186],8.92,[207,186,186,238,188,186,188,188,186],2.47,[207,186,197,240,188,186,188,188,186],3.69,[207,186,200,242,188,186,188,188,186],29.73,[207,186,190,244,188,186,188,188,186],1.11,[207,186,191,246,188,186,188,188,186],2.42,[207,186,207,248,188,186,188,188,186],7.89,[115,186,186,250,188,186,188,188,186],2.4,[115,186,190,252,188,186,188,188,186],1.7,[115,186,191,250,188,186,188,188,186],[255,186,186,256,188,186,188,188,186],7,1.5,[255,186,200,258,188,186,188,188,186],4.64,[255,186,190,260,188,186,188,188,186],0.6,[255,186,191,262,188,186,188,188,186],1.3,[264,186,186,265,188,186,188,188,186],8,3.79,[264,186,197,267,188,186,188,188,186],3.38,[264,186,200,269,188,186,188,188,186],7.03,[264,186,190,271,188,186,188,188,186],1.62,[264,186,191,273,188,186,188,188,186],3.41,[264,186,207,275,188,186,188,188,186],3.85,[277,186,186,252,188,186,188,188,186],9,[277,186,190,279,188,186,188,188,186],1.1,[277,186,191,252,188,186,188,188,186],[282,186,186,283,188,186,188,188,186],10,3.7,[282,186,197,285,188,186,188,188,186],7.1,[282,186,200,287,188,186,188,188,186],7.5,[282,186,190,289,188,186,188,188,186],2.9,[282,186,191,197,188,186,188,188,186],[282,186,207,292,188,186,188,188,186],4.84,[294,186,186,295,188,186,188,188,186],11,1.6,[294,186,197,297,188,186,188,188,186],2.2,[294,186,200,299,188,186,188,188,186],4.7,[294,186,190,301,188,186,188,188,186],0.4,[294,186,191,190,188,186,188,188,186],[294,186,207,304,188,186,188,188,186],1.98,[306,186,186,307,188,186,188,188,186],12,2.53,[306,186,197,309,188,186,188,188,186],6.83,[306,186,200,311,188,186,188,188,186],21.49,[306,186,190,313,188,186,188,188,186],1.17,[306,186,191,315,188,186,188,188,186],2.52,[306,186,207,317,188,186,188,188,186],6.91,[319,186,186,295,188,186,188,188,186],13,[319,186,190,279,188,186,188,188,186],[319,186,191,297,188,186,188,188,186],[323,186,186,256,188,186,188,188,186],14,[323,186,190,260,188,186,188,188,186],[323,186,191,326,188,186,188,188,186],0.9,[328,186,186,295,188,186,188,188,186],15,[328,186,200,330,188,186,188,188,186],5.1,[328,186,191,332,188,186,188,188,186],3.1,[],[120],[],[],[338],"ATE RMSE (cm) on TUM-RGBD, trajectories aligned with Horn's closed-form solution before ATE (App. C; whether scale was estimated is not stated); average of three runs unless noted; baseline numbers taken from the respective papers where available, otherwise reproduced; top block dense neural RGB-D methods, bottom block traditional dense and sparse SLAM; LC = loop closure. N\u002FA cells (not reported) omitted.",{"slug":340,"group":341,"sourceId":5,"sourceLabel":6,"table":342,"selfRows":200,"metrics":343,"seqs":354,"entrants":359,"cells":369,"outcomes":420,"locators":421,"hardware":423,"wordings":424,"notes":425},"loopyslam2024-table-11","loopyslam2024:Table 11","Table 11",[344,347,350,352],{"label":345,"unit":124,"statistic":346,"alignment":126},"Depth L1 on mesh at random poses","mean",{"label":348,"unit":349,"statistic":126,"alignment":126},"Precision at 1 cm","%",{"label":351,"unit":349,"statistic":126,"alignment":126},"Recall at 1 cm",{"label":353,"unit":349,"statistic":126,"alignment":126},"F1 at 1 cm",[355],{"dataset":356,"sequence":357,"environment":358},"Replica","Avg. of 8 scenes","synthetic indoor rooms and offices",[360,361,362,363,364,366,367],{"name":145,"methodId":146,"linkable":147,"proposed":78,"self":78},{"name":149,"methodId":87,"linkable":78,"proposed":78,"self":78},{"name":156,"methodId":157,"linkable":147,"proposed":78,"self":78},{"name":159,"methodId":160,"linkable":147,"proposed":78,"self":78},{"name":365,"methodId":87,"linkable":78,"proposed":78,"self":78},"GO-SLAM [76] (reproduced, random poses)",{"name":153,"methodId":154,"linkable":147,"proposed":78,"self":78},{"name":368,"methodId":5,"linkable":147,"proposed":147,"self":147},"Loopy-SLAM (Ours)",[370,372,374,376,378,380,382,384,386,388,390,392,394,396,398,400,402,404,406,408,410,412,414,416,418],[186,186,186,371,188,186,188,188,186],2.97,[186,190,186,373,188,186,188,188,186],44.1,[186,191,186,375,188,186,188,188,186],43.69,[186,197,186,377,188,186,188,188,186],43.86,[190,186,186,379,188,186,188,188,186],2.46,[190,190,186,381,188,186,188,188,186],55.73,[190,191,186,383,188,186,188,188,186],49.13,[190,197,186,385,188,186,188,188,186],52.2,[191,186,186,387,188,186,188,188,186],1.18,[191,190,186,389,188,186,188,188,186],74.25,[191,191,186,391,188,186,188,188,186],84.79,[191,197,186,393,188,186,188,188,186],79.09,[197,186,186,395,188,186,188,188,186],1.51,[200,186,186,397,188,186,188,188,186],4.68,[200,190,186,399,188,186,188,188,186],30.73,[200,191,186,401,188,186,188,188,186],23.04,[200,197,186,403,188,186,188,188,186],26.34,[207,186,186,405,188,186,188,188,186],0.44,[207,190,186,407,188,186,188,188,186],96.99,[207,191,186,409,188,186,188,188,186],83.59,[207,197,186,411,188,186,188,188,186],89.77,[115,186,186,413,188,186,188,188,186],0.35,[115,190,186,415,188,186,188,188,186],98.53,[115,191,186,417,188,186,188,188,186],84.18,[115,197,186,419,188,186,188,188,186],90.77,[],[422],"Table 11; Fig. 3a",[],[],[426],"Replica mesh reconstruction averaged over 8 scenes: meshes from marching cubes, ICP-aligned to ground truth before precision and recall; precision, recall and F1 at a 1 cm threshold; depth L1 renders depth from 1000 random viewpoints on reconstructed and ground-truth meshes. GO-SLAM depth L1 3.38 is the value GO-SLAM reports with ground-truth poses; the starred 4.68 is the authors' reproduction from random poses. Truncated to scene averages.",{"slug":428,"group":429,"sourceId":5,"sourceLabel":6,"table":430,"selfRows":197,"metrics":431,"seqs":433,"entrants":442,"cells":453,"outcomes":484,"locators":485,"hardware":486,"wordings":487,"notes":488},"loopyslam2024-table-3","loopyslam2024:Table 3","Table 3",[432],{"label":123,"unit":124,"statistic":125,"alignment":126},[434,438,440],{"dataset":435,"sequence":436,"environment":437},"ScanNet","Avg.-6 (scenes 00, 59, 106, 169, 181, 207)","ScanNet real-world RGBD scenes; reference poses from BundleFusion; scene 54 is the only multi-room and the largest scene",{"dataset":435,"sequence":439,"environment":437},"scene 54",{"dataset":435,"sequence":441,"environment":437},"Avg.-9 (adds 54, 233, 465)",[443,444,445,447,448,449,450,452],{"name":149,"methodId":87,"linkable":78,"proposed":78,"self":78},{"name":159,"methodId":160,"linkable":147,"proposed":78,"self":78},{"name":446,"methodId":87,"linkable":78,"proposed":78,"self":78},"MIPS-Fusion [57]",{"name":145,"methodId":146,"linkable":147,"proposed":78,"self":78},{"name":156,"methodId":157,"linkable":147,"proposed":78,"self":78},{"name":153,"methodId":154,"linkable":147,"proposed":78,"self":78},{"name":451,"methodId":87,"linkable":78,"proposed":78,"self":78},"GO-SLAM [76]",{"name":368,"methodId":5,"linkable":147,"proposed":147,"self":147},[454,456,458,459,461,463,464,466,468,470,472,474,476,477,479,480,481,483],[186,186,186,455,188,186,188,188,186],18.5,[190,186,186,457,188,186,188,188,186],8.8,[191,186,186,282,188,186,188,188,186],[197,186,190,460,188,186,188,188,186],20.9,[197,186,186,462,188,186,188,188,186],10.7,[197,186,191,319,188,186,188,188,186],[200,186,190,465,188,186,188,188,186],36.3,[200,186,186,467,188,186,188,188,186],7.4,[200,186,191,469,188,186,188,188,186],11.3,[207,186,190,471,188,186,188,188,186],28,[207,186,186,473,188,186,188,188,186],12.2,[207,186,191,475,188,186,188,188,186],14.3,[115,186,190,457,188,186,188,188,186],[115,186,186,478,188,186,188,188,186],6.9,[115,186,191,255,188,186,188,188,186],[255,186,190,287,188,186,188,188,186],[255,186,186,482,188,186,188,188,186],7.7,[255,186,191,482,188,186,188,188,186],[],[430],[],[],[489],"ATE RMSE (cm) on ScanNet after Horn closed-form alignment (App. C); reference poses come from BundleFusion; Avg.-6 and Avg.-9 average over 6 and 9 scenes. Truncated: only scene 54 (the only multi-room and largest scene) and the two averages are kept; '-' cells omitted.",{"slug":491,"group":492,"sourceId":5,"sourceLabel":6,"table":493,"selfRows":197,"metrics":494,"seqs":503,"entrants":507,"cells":513,"outcomes":541,"locators":542,"hardware":543,"wordings":545,"notes":546},"loopyslam2024-table-5","loopyslam2024:Table 5","Table 5",[495,498,500],{"label":496,"unit":497,"statistic":126,"alignment":46},"Tracking\u002FFrame","s",{"label":499,"unit":497,"statistic":126,"alignment":46},"Mapping\u002FFrame",{"label":501,"unit":502,"statistic":126,"alignment":46},"Embedding Size (total memory of the map representation)","MB",[504],{"dataset":356,"sequence":505,"environment":506},"office 0","synthetic office",[508,509,510,511,512],{"name":145,"methodId":146,"linkable":147,"proposed":78,"self":78},{"name":149,"methodId":87,"linkable":78,"proposed":78,"self":78},{"name":153,"methodId":154,"linkable":147,"proposed":78,"self":78},{"name":156,"methodId":157,"linkable":147,"proposed":78,"self":78},{"name":368,"methodId":5,"linkable":147,"proposed":147,"self":147},[514,516,518,520,522,524,526,528,530,532,534,535,537,538,539],[186,186,186,515,188,186,188,188,186],1.32,[186,190,186,517,188,186,188,188,186],10.92,[186,191,186,519,188,186,188,188,186],95.86,[190,186,186,521,188,186,188,188,186],0.36,[190,190,186,523,188,186,188,188,186],0.55,[190,191,186,525,188,186,188,188,186],0.149,[191,186,186,527,188,186,188,188,186],0.85,[191,190,186,529,188,186,188,188,186],9.85,[191,191,186,531,188,186,188,188,186],27.23,[197,186,186,533,188,186,188,188,186],0.12,[197,190,186,405,188,186,188,188,186],[197,191,186,536,188,186,188,188,186],45.46,[200,186,186,527,188,186,186,188,186],[200,190,186,529,188,186,186,188,186],[200,191,186,540,188,186,188,188,186],60.92,[],[493],[544],"NVIDIA GPU with at most 12 GB memory (model not stated)",[],[547],"Runtime and memory on Replica office 0. Loopy-SLAM tracking and mapping times are identical to the Point-SLAM row (the text says they are equivalent excluding loop closure). GO-SLAM reports a single 0.125 s value spanning both per-frame columns. Per-iteration columns not extracted.",[549,554],{"group":550,"slug":551,"sourceLabel":6,"table":552,"selfRows":191,"datasets":553},"loopyslam2024:Text Sec.4.4","loopyslam2024-text-sec-4-4","Text Sec.4.4",[92],{"group":555,"slug":556,"sourceLabel":6,"table":557,"selfRows":190,"datasets":558},"loopyslam2024:Table 1","loopyslam2024-table-1","Table 1",[356],1790510665201]