Pith. sign in

REVIEW 3 major objections 4 minor 111 references

Old elevation maps replace feature matching for real-time terrain depth

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · glm-5.2

2026-07-10 02:31 UTC pith:47POUFZN

load-bearing objection Reasonable framework for wildfire terrain mapping, but the central claims are empirically unsupported — no quantitative depth metrics, no runtime numbers, baselines dismissed qualitatively. the 3 major comments →

arxiv 2607.08711 v1 pith:47POUFZN submitted 2026-07-09 cs.CV cs.LG

LTM: Large-scale Terrain Model for Wildfire-prone Landscapes

Xiao Fu , Yue Hu , Meida Chen , Peter Anthony Beerel , Barath Raghavan This is my paper
classification cs.CV cs.LG
keywords reconstructionimagesterrainwildfire-pronecomputationaldatademsexpensive
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that for wildfire-prone landscapes — vast, vegetated, and poorly served by existing 3D reconstruction methods — the right 3D representation is not a point cloud, mesh, or neural radiance field, but a raster Digital Elevation Model (DEM): a 2D grid where each cell stores one elevation value. The author's case rests on three claims. First, raster DEMs scale better than point-based representations for large areas (quadratic versus cubic in width), making them practical for 10 km × 10 km wildlands. Second, because natural terrain is structurally stable relative to urban scenes, an outdated DEM can serve as a geometric prior for current imagery: a physics-based ray-tracing algorithm casts a ray from the camera through each image pixel, steps through DEM grid cells, and stops when the ray intersects the stored terrain elevation, producing both a depth map and a direct image-to-DEM correspondence without any feature matching — the step that dominates cost in conventional pipelines. Third, by fusing this DEM-constrained depth with neural monocular depth estimation and semantic segmentation of vegetation types, the method produces updated terrain models and fuel maps (categorizing vegetation as grass, shrub, or tree) at resolutions useful for wildfire behavior simulators. The author validates the approach using a custom simulator built on real DEM data from the Getty Fire site, and reports that baseline monocular depth methods produce errors one to two orders of magnitude larger than the proposed method on mountainous terrain, while the proposed method maintains depth errors within tens of meters.

Core claim

The paper's central mechanism is a ray-tracing algorithm that aligns each pixel of a posed ground-level image to a cell in a raster Digital Elevation Model — a 2D grid where each cell stores terrain elevation. By casting a ray from the camera through each image pixel and stepping through DEM grid cells until the ray's elevation drops below the terrain surface stored in the DEM, the method produces both a depth map and a direct image-to-DEM correspondence without any feature matching. This replaces the computationally dominant step of conventional 3D reconstruction (pairwise feature matching, which scales as the square of the number of images) with a per-pixel operation that scales linearly.

What carries the argument

Pixel-based on-raster ray tracing (Algorithm 1): given camera pose, location, field of view, and a DEM grid, for each image pixel a ray is generated and marched through DEM cells until the ray elevation falls below the terrain elevation, yielding both a depth value and a DEM coordinate for that pixel.

Load-bearing premise

The method assumes that an outdated DEM accurately represents the current terrain geometry, because the ray-tracing algorithm traces rays against the old elevation grid. But the paper's own problem statement acknowledges that terrain changes due to prescribed burns, vegetation growth, and habitat shifts — the very changes that motivate updating the DEM. The paper does not quantify how much terrain change the method tolerates before depth estimates become unreliable.

What would settle it

If depth errors from the ray-traced DEM alignment do not stay within tens of meters on real mountainous terrain, or if the method's runtime does not achieve real-time performance on standard hardware for 10 km × 10 km areas, the central efficiency and accuracy claims would be undermined.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The manuscript proposes a multi-modal 3D terrain mapping framework for wildfire-prone landscapes. The core idea is to use outdated Digital Elevation Models (DEMs) as geometric priors for image-based 3D reconstruction. The authors introduce a physics-based pixel-to-pixel alignment algorithm (Algorithm 1) that uses ray tracing to map image pixels to DEM raster coordinates, thereby eliminating the need for expensive feature matching. Additionally, the authors develop a vegetated terrain simulator using Unreal Engine to generate photorealistic images with ground-truth geometry and semantics. The framework is evaluated on a real-world site (the Getty Fire area) and in simulation, demonstrating an end-to-end pipeline from image acquisition to 3D semantic fuel mapping.

Significance. The problem addressed is highly relevant to environmental monitoring and disaster management. The strategic use of raster-based DEMs as a 3D representation for large-scale, vegetated environments is well-motivated, and the computational complexity analysis (Table 1) arguing for the scalability of raster-based methods over point clouds is a notable strength. The development of a custom simulator for vegetated terrains fills a gap in existing datasets, which are predominantly urban. However, the empirical validation of the central claims is currently insufficient.

major comments (3)
  1. The central claim of 'significant improvements in reconstruction accuracy and computational efficiency over existing techniques' (Abstract) is not supported by quantitative evidence. The Evaluation Metrics section names RMSE as the depth estimation metric, but no RMSE values are reported anywhere in the manuscript. The depth estimation results (Figure 5) are purely qualitative, and the text explicitly states that 'Baseline methods are omitted from quantitative comparison due to depth predictions exhibiting 1-2 orders of magnitude greater error.' This makes it impossible to verify the claimed improvements. The authors must report actual RMSE values for their method (TopoDepth) and the baselines (UniDepth, DepthPro) on their dataset. Without these numbers, the core claim is empirically unsupported.
  2. The claim of 'real-time performance' and 'computational efficiency' (Abstract) is not backed by any runtime measurements. The only support is the theoretical big-O complexity analysis in Table 1. While the theoretical argument for raster-based scaling is sound, the specific claim of real-time performance requires empirical runtime data (e.g., processing time per image or per km²) on the specified hardware (NVIDIA RTX 3090 / A6000).
  3. The paper's own Problem Formulation states that 'Terrain models are frequently outdated due to natural environmental changes, including prescribed burns, vegetation growth, and habitat shifts.' This creates a tension with the core assumption that outdated DEMs serve as reliable geometric priors for current terrain (Overview: 'This stability enables historical DEMs to serve as reliable geometric priors'). If the terrain has changed, Algorithm 1's ray-tracing will produce incorrect depth estimates. The paper does not quantify how much terrain change the method tolerates before depth estimates degrade significantly. A sensitivity analysis or discussion of this limitation is needed.
minor comments (4)
  1. The depth estimation approach explicitly follows the authors' own prior work, FireLoc (Fu et al. 2024): 'Following FireLoc (Fu et al. 2024), we fuse neural depth estimation with DEM constraints through pixel-wise sampling and RANSAC regression.' The manuscript should clearly delineate what is novel in this paper versus what is inherited from FireLoc.
  2. Table 2 reports sim-to-real image similarity metrics (SSIM=0.36, LPIPS=0.44). An SSIM of 0.36 is quite low, suggesting poor structural similarity between simulation and reality. The text claims 'strong sim-to-real fidelity,' which seems inconsistent with these numbers. Please clarify or contextualize these metrics.
  3. Algorithm 1 uses variables ΔX, ΔY, and Δelev for ray traversal, but the step size or method for determining these deltas is not specified. This affects the reproducibility of the ray-tracing approach.
  4. The paper mentions that 'Performance degrades under significant occlusion conditions' for the depth map with terrain priors. It would be helpful to quantify or show examples of this degradation, as vegetation occlusion is a primary challenge in the targeted environments.

Circularity Check

0 steps flagged

No significant circularity; one descriptive self-citation to FireLoc that is not load-bearing for any mathematical claim.

full rationale

The paper's core algorithm (Algorithm 1, pixel-based on-raster ray tracing) is a straightforward geometric procedure: given camera pose and DEM, it traces rays through the DEM raster to produce a depth map and image-DEM alignment. This is independently derivable from basic projective geometry and is not circular. The depth map output does contain DEM information by construction (it is a rendering of the DEM from the camera viewpoint), but the paper is transparent that the DEM is an input prior, not a derived quantity. The fusion with neural depth estimation ('Following FireLoc (Fu et al. 2024), we fuse neural depth estimation with DEM constraints through pixel-wise sampling and RANSAC regression') is a self-citation with 4/5 author overlap, but it describes a standard RANSAC fusion technique rather than invoking an unverified theorem or uniqueness result. No 'prediction' reduces to a fitted input by construction. The complexity analysis (Table 1) is a theoretical big-O comparison, not a circular derivation. The absence of quantitative depth metrics (RMSE named but never reported) is a serious empirical support problem, but it is a correctness/evaluation issue, not circularity. The derivation chain is self-contained against external benchmarks in principle, and the self-citation is descriptive rather than load-bearing.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 1 invented entities

The paper introduces one main invented entity (the simulator) and relies on several domain assumptions, the most fragile being the DEM reliability assumption which conflicts with the paper's own motivation.

free parameters (4)
  • Ray step size (ΔX, ΔY, Δelev) = unspecified
    Algorithm 1 uses incremental ray traversal with step sizes ΔX, ΔY, Δelev that are not specified. These affect both accuracy and computational cost.
  • RANSAC regression parameters = unspecified
    The depth fusion step uses RANSAC regression following FireLoc, but threshold and iteration parameters are not stated.
  • Camera relative height offset = tripod height (example)
    Algorithm 1 line 3 uses 'common camera relative height offset (e.g., tripod height)' as initial elevation when barometer data is unavailable. This is a hand-set parameter.
  • Fuel map resolution s = 30 m
    The fuel map spatial resolution is set to 30m following LANDFIRE standards, not derived from the method.
axioms (4)
  • domain assumption Historical DEMs serve as reliable geometric priors for current terrain
    Overview section: 'This stability enables historical DEMs to serve as reliable geometric priors for terrain reconstruction using current imagery.' This is the load-bearing assumption that conflicts with the paper's own motivation about terrain change.
  • domain assumption Raster-based DEM representation is sufficient for wildfire terrain (no overhangs/caves needed)
    Method section: 'the raster 3D representation cannot capture hollow and concave features of landscape (e.g., overhangs, caves). But those scenarios are less relevant to wildfire propagation.'
  • domain assumption Monocular depth estimation networks can capture surface variations not in the DEM
    Depth map section: 'This hybrid method leverages DEMs for stable terrain baselines while neural networks capture detailed surface variations.' No evidence is provided that neural depth adds meaningful accuracy beyond the DEM prior in these scenes.
  • ad hoc to paper The simulator accurately represents real-world wildfire terrain challenges
    The entire quantitative evaluation rests on the simulator. Table 2 shows LPIPS=0.4437 and SSIM=0.3614 for sim-to-real, which are moderate at best, yet the simulator is used as the primary evaluation tool.
invented entities (1)
  • Vegetated perturbation topographical simulator no independent evidence
    purpose: Generate controlled wildfire-prone terrain imagery with ground truth for evaluation
    The simulator is a new tool built on Unreal Engine. It is not independently validated against real-world performance beyond Table 2's moderate sim-to-real metrics. No code is released for external verification.

pith-pipeline@v1.1.0-glm · 16563 in / 4103 out tokens · 191308 ms · 2026-07-10T02:31:42.312849+00:00 · methodology

0 comments
read the original abstract

Accurate 3D terrain maps are essential for emergency response when assessing wildfire hazards. However, wildfire-prone regions often span vast areas where conventional reconstruction methods underperform. Airborne LiDAR systems provide high-resolution terrain data, but they are expensive and infrequently updated. Image-based methods offer a lower-cost alternative, but struggle due to sparse visual features and limited image overlap. We propose a multi-modal reconstruction framework leveraging outdated Digital Elevation Models (DEMs) as geometric priors for image-based 3D reconstruction. Our key innovation is physics-based pixel-pixel alignment between images and DEM data, dramatically reducing computational complexity by eliminating expensive feature matching procedures. To validate our approach, we developed a large-terrain simulator based on a real wildfire-prone area, generating realistic images enabling a comprehensive evaluation. Given posed images and legacy DEMs, our method produces high-fidelity depth maps while maintaining real-time performance. We find significant improvements in reconstruction accuracy and computational efficiency over existing techniques, offering a scalable solution for wildfire response.

Figures

Figures reproduced from arXiv: 2607.08711 by Barath Raghavan, Meida Chen, Peter Anthony Beerel, Xiao Fu, Yue Hu.

Figure 1
Figure 1. Figure 1: A novel framework for large-scale dynamic terrain [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: DEM efficiently represents large-scale 3D terrain [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: We present 2 pairs of real-world, simulated images, and their respective image-based segmentation results. The [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Fuel maps comparing real-world and simulated [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Monocular depth estimation in mountainous terrain. Our method outperforms UniDepth and DepthPro, which struggle [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

111 extracted references · 111 canonical work pages · 7 internal anchors

  1. [1]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

    work page

  2. [2]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

    work page

  3. [3]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    work page 1980

  4. [4]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

    work page

  5. [5]

    FirstName LastName , title =

    work page

  6. [6]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Dense depth priors for neural radiance fields from sparse input views , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  7. [7]

    Pixelwise View Selection for Unstructured Multi-View Stereo , booktitle=

    Sch\". Pixelwise View Selection for Unstructured Multi-View Stereo , booktitle=

    work page

  8. [8]

    Structure-from-Motion Revisited , booktitle=

    Sch\". Structure-from-Motion Revisited , booktitle=

    work page

  9. [9]

    Communications of the ACM , volume=

    Nerf: Representing scenes as neural radiance fields for view synthesis , author=. Communications of the ACM , volume=. 2021 , publisher=

    work page 2021

  10. [10]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Neural Scene Chronology , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  11. [11]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Superglue: Learning feature matching with graph neural networks , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  12. [12]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Change-Aware Sampling and Contrastive Learning for Satellite Images , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  13. [13]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Privacy-Preserving Representations are not Enough: Recovering Scene Content from Camera Poses , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  14. [14]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Visual saliency transformer , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  15. [15]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    End-to-end pseudo-lidar for image-based 3d object detection , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  16. [16]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Autosdf: Shape priors for 3d completion, reconstruction and generation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  17. [17]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Gs3d: An efficient 3d object detection framework for autonomous driving , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  18. [18]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    Geonet: Geometric neural network for joint depth and surface normal estimation , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    work page

  19. [19]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Marr revisited: 2d-3d alignment via surface normal prediction , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  20. [20]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Deepfusion: Lidar-camera deep fusion for multi-modal 3d object detection , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  21. [21]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Persistent Nature: A Generative Model of Unbounded 3D Worlds , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  22. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    DINER: Depth-aware Image-based NEural Radiance fields , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  23. [23]

    2008 ieee conference on computer vision and pattern recognition , pages=

    Im2gps: estimating geographic information from a single image , author=. 2008 ieee conference on computer vision and pattern recognition , pages=. 2008 , organization=

    work page 2008

  24. [24]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Pseudo-lidar from visual depth estimation: Bridging the gap in 3d object detection for autonomous driving , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  25. [25]

    Computer Vision--ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part II 12 , pages=

    Large scale visual geo-localization of images in mountainous terrain , author=. Computer Vision--ECCV 2012: 12th European Conference on Computer Vision, Florence, Italy, October 7-13, 2012, Proceedings, Part II 12 , pages=. 2012 , organization=

    work page 2012

  26. [26]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Megadepth: Learning single-view depth prediction from internet photos , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  27. [27]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Digging into self-supervised monocular depth estimation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  28. [28]

    FirstName Alpher , title =

    work page

  29. [29]

    Journal of Foo , volume = 13, number = 1, pages =

    FirstName Alpher and FirstName Fotheringham-Smythe , title =. Journal of Foo , volume = 13, number = 1, pages =

    work page

  30. [30]

    IEEE transactions on robotics , volume=

    ORB-SLAM: a versatile and accurate monocular SLAM system , author=. IEEE transactions on robotics , volume=. 2015 , publisher=

    work page 2015

  31. [31]

    Journal of Foo , volume = 14, number = 1, pages =

    FirstName Alpher and FirstName Fotheringham-Smythe and FirstName Gamow , title =. Journal of Foo , volume = 14, number = 1, pages =

    work page

  32. [32]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Self-Supervised Object Detection from Egocentric Videos , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  33. [33]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Virtual Sparse Convolution for Multimodal 3D Object Detection , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  34. [34]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Mega-nerf: Scalable construction of large-scale nerfs for virtual fly-throughs , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  35. [35]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Sat-nerf: Learning multi-view satellite photogrammetry with transient objects and shadow modeling using rpc cameras , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  36. [36]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Occupancy networks: Learning 3d reconstruction in function space , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  37. [37]

    FirstName Alpher and FirstName Gamow , title =

    work page

  38. [38]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Octnet: Learning deep 3d representations at high resolutions , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  39. [39]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Bridge 2D-3D: Uncertainty-aware Hierarchical Registration Network with Domain Alignment , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  40. [40]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Zero-shot scene change detection , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  41. [41]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    LoFTR: Detector-free local feature matching with transformers , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  42. [42]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  43. [43]

    European Conference on Computer Vision , pages=

    Meshloc: Mesh-based visual localization , author=. European Conference on Computer Vision , pages=. 2022 , organization=

    work page 2022

  44. [44]

    2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Monocular camera localization in prior lidar maps with 2d-3d line correspondences , author=. 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2020 , organization=

    work page 2020

  45. [45]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    DeepI2P: Image-to-point cloud registration via deep classification , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  46. [46]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Accelerated coordinate encoding: Learning to relocalize in minutes using rgb and poses , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  47. [47]

    2024 , eprint=

    DepthFM: Fast Monocular Depth Estimation with Flow Matching , author=. 2024 , eprint=

    work page 2024

  48. [48]

    Advances in Neural Information Processing Systems , volume=

    Differentiable registration of images and lidar point clouds with voxelpoint-to-pixel matching , author=. Advances in Neural Information Processing Systems , volume=

    work page

  49. [49]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Map-relative pose regression for visual re-localization , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  50. [50]

    Natural hazards , volume=

    Avalanche hazard mapping over large undocumented areas , author=. Natural hazards , volume=. 2011 , publisher=

    work page 2011

  51. [51]

    International Journal of Remote Sensing , volume=

    SRTM DEM and its application advances , author=. International Journal of Remote Sensing , volume=. 2011 , publisher=

    work page 2011

  52. [52]

    Mapping volcanic terrain using high-resolution and 3D satellite remote sensing , author=

    work page

  53. [53]

    Current Forestry Reports , volume=

    Structure from motion photogrammetry in forestry: A review , author=. Current Forestry Reports , volume=. 2019 , publisher=

    work page 2019

  54. [54]

    Conference on robot learning , pages=

    CARLA: An open urban driving simulator , author=. Conference on robot learning , pages=. 2017 , organization=

    work page 2017

  55. [55]

    2013 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) , pages=

    Real-time RGB-D camera relocalization , author=. 2013 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) , pages=. 2013 , organization=

    work page 2013

  56. [56]

    The international journal of robotics research , volume=

    Vision meets robotics: The kitti dataset , author=. The international journal of robotics research , volume=. 2013 , publisher=

    work page 2013

  57. [57]

    ROVER: A Multiseason Dataset for Visual SLAM , year=

    Schmidt, Fabian and Daubermann, Julian and Mitschke, Marcel and Blessing, Constantin and Meyer, Stephan and Enzweiler, Markus and Valada, Abhinav , journal=. ROVER: A Multiseason Dataset for Visual SLAM , year=

    work page

  58. [58]

    Semantic Feature Matching for Robust Mapping in Agriculture

    Semantic feature matching for robust mapping in agriculture , author=. arXiv preprint arXiv:2107.04178 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  59. [59]

    Piccinelli, Luigi and Yang, Yung-Hsu and Sakaridis, Christos and Segu, Mattia and Li, Siyuan and Van Gool, Luc and Yu, Fisher , booktitle =

    work page

  60. [60]

    ECCV , year =

    Grounding Image Matching in 3D with MASt3R , author=. ECCV , year =

    work page

  61. [61]

    Depth Anything V2

    Depth Anything V2 , author=. arXiv:2406.09414 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  62. [62]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Metric3D: Towards zero-shot metric 3d prediction from a single image , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  63. [63]

    The Rothermel surface fire spread model and associated developments: A comprehensive explanation , author=. Gen. Tech. Rep. RMRS-GTR-371. Fort Collins, CO: US Department of Agriculture, Forest Service, Rocky Mountain Research Station. 121 p. , volume=

    work page

  64. [64]

    The Bark Beetles, Fuels, and Fire Bibliography , pages=

    Aids to determining fuel models for estimating fire behavior , author=. The Bark Beetles, Fuels, and Fire Bibliography , pages=

    work page

  65. [65]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Masked-attention mask transformer for universal image segmentation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  66. [66]

    Marigold: Affordable Adaptation of Diffusion-Based Image Generators for Image Analysis

    Marigold: Affordable Adaptation of Diffusion-Based Image Generators for Image Analysis , author=. arXiv preprint arXiv:2505.09358 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  67. [67]

    Depth Pro: Sharp Monocular Metric Depth in Less Than a Second

    Depth pro: Sharp monocular metric depth in less than a second , author=. arXiv preprint arXiv:2410.02073 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  68. [68]

    Metric3Dv2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth and Surface Normal Estimation

    Hu, Mu and Yin, Wei and Zhang, Chi and Cai, Zhipeng and Long, Xiaoxiao and Chen, Hao and Wang, Kaixuan and Yu, Gang and Shen, Chunhua and Shen, Shaojie , title=. arXiv preprint arXiv:2404.15506 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  69. [69]

    CVPR , year=

    Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data , author=. CVPR , year=

    work page

  70. [70]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Random Mapping Method for Large-Scale Terrain Modeling , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

  71. [71]

    CVPR , year =

    DUSt3R: Geometric 3D Vision Made Easy , author=. CVPR , year =

    work page

  72. [72]

    2024 , eprint=

    Large Spatial Model: End-to-end Unposed Images to Semantic 3D , author=. 2024 , eprint=

    work page 2024

  73. [73]

    International Conference on Learning Representations , year=

    Language-driven Semantic Segmentation , author=. International Conference on Learning Representations , year=

    work page

  74. [74]

    2023 , eprint=

    DUSt3R: Geometric 3D Vision Made Easy , author=. 2023 , eprint=

    work page 2023

  75. [75]

    UniDepthV2: Universal Monocular Metric Depth Estimation Made Simpler

    Luigi Piccinelli and Christos Sakaridis and Yung-Hsu Yang and Mattia Segu and Siyuan Li and Wim Abbeloos and Luc Van Gool , year=. 2502.20110 , archivePrefix=

    work page internal anchor Pith review Pith/arXiv arXiv

  76. [76]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Pow3r: Empowering unconstrained 3d reconstruction with camera and scene priors , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    work page

  77. [77]

    Proceedings of the IEEE international conference on computer vision , pages=

    Posenet: A convolutional network for real-time 6-dof camera relocalization , author=. Proceedings of the IEEE international conference on computer vision , pages=

    work page

  78. [78]

    Remote Sensing , volume=

    Leveraging Neural Radiance Fields for Large-Scale 3D Reconstruction from Aerial Imagery , author=. Remote Sensing , volume=. 2024 , publisher=

    work page 2024

  79. [79]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    A large-scale outdoor multi-modal dataset and benchmark for novel view synthesis and implicit scene reconstruction , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  80. [80]

    GigaSLAM: Large-Scale Monocular SLAM with Hierarchical Gaussian Splats

    GigaSLAM: Large-Scale Monocular SLAM with Hierarchical Gaussian Splats , author=. arXiv preprint arXiv:2503.08071 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

Showing first 80 references.