Papers for

mapping software developers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Large language models improve place descriptions with spatial generalization

From Geographic Memory to Spatial Generalization in Large Language Models

Abstract: Large language models (LLMs) can associate places with geographic descriptions, but recovering an observed association does not establish an ability to generalize across geographic space. We distinguish geospatial memory, the recovery of location--description associations observed during training, from spatial generalization, the prediction of descriptions for locations in unobserved regions. We investigate whether continuous location representations improve this generalization and encourage geographic knowledge to become spatially organized within an LLM. We frame this organization as a latent spatial field: a learned mapping from locations to descriptions connected through spatial dependencies. Building on a continuous spatial encoder, we project location representations into the model's embedding space and supply them as spatial tokens. We evaluate the resulting models on two newly curated geo-text benchmarks for Los Angeles and London using two complementary assessments. Geospatial captioning tests both memory and spatial generalization by generating descriptions at observed locations and at locations beyond the spatial coverage of the training observations. A separate geospatial grounding model tests whether generated location--description pairs provide useful additional supervision for mapping text to geographic space. Across multiple LLM families, continuous location representations improve captioning at unobserved locations relative to textual coordinates, while their generated descriptions yield modest but consistent improvements in downstream grounding. Layer-wise probing and representation clustering provide additional evidence consistent with spatially organized geographic representations. Together, these findings show the value of evaluating spatial generalization separately from memory and of incorporating continuous spatial representations into language models.

Thu 19 NovComputation and LanguageArtificial IntelligenceMachine Learning
The gist
This paper looks at how language models, like AI text generators, understand and describe places. The authors show that simply remembering associations between places and descriptions isn’t enough; models need to generalize descriptions to locations they haven't seen before. By giving the models continuous location information instead of just text names, their ability to describe new places improves. The authors test this on data from Los Angeles and London and show that this approach helps the models organize geographical knowledge better.
Open → 2611.25703v1

Task specific vision language models generate paths on remote sensing images

RemoteVLPG: How to Obtain a Specific VLM for Language-Guided Path Generation on Remote Sensing Images?

Abstract: We study how to obtain a task-specific Vision-Language Model (VLM) for language-guided path generation between start and end points on remote sensing images. We explore this question from three aspects. First, we implement three widely used architectural approaches and compare their performance. Second, we design a scalable method for constructing pre-training data and build an effective pre-training dataset with a degree of diversity. Methods trained on this dataset achieve promising zero-shot performance. Third, we construct three benchmarks that support stratified evaluation and design three metrics for path generation. We investigate this problem through extensive experiments. The results show that the zero-shot performance of methods trained on the constructed pre-training dataset can exceed that of some advanced commercial VLMs. Code and Dataset can be available at https://github.com/jzdang/RemoteVLPG

Thu 19 NovComputer Vision and Pattern Recognition
The gist
Generating paths between two points on aerial or satellite images is challenging because it needs understanding both images and language instructions. The authors studied how to create specialized vision-language models that follow language instructions to draw paths on these images. They built a diverse dataset and tested different model designs, showing their approach works well even without extra training on specific tasks. Their models can outperform some commercial systems in zero-shot scenarios, meaning they work without specialized retraining.
Open → 2611.25458v1

Unified framework improves loop closure detection in visual mapping

UniLoop: Unified Loop Closure Detection from Retrieval to Geometric Verification and Relative Pose Estimation

Abstract: Loop closure detection conventionally cascades separate models for global retrieval, geometric verification, and relative pose estimation, leading to redundant computation and disconnected visual representations. We propose UniLoop, a unified framework that reuses the hierarchical representations of the pretrained geometric foundation model throughout this pipeline. A shared image encoder produces features for compact global descriptors; the same features are passed to the cross-view decoder for pairwise verification; and the original geometric predictions provide relative camera constraints for pose-graph optimization. Experiments on place recognition, loop verification, and relative pose estimation show that UniLoop improves loop-closure robustness and accuracy while reducing redundant backbone computation relative to modular pipelines. UniLoop achieves the best Recall@10 on MSLS (96.6) and Pitts30k (99.3) among the compared methods, while its verification branch reaches 95.63 AP and 96.29 ROC-AUC, surpassing the existing geometric verification baselines.

Wed 4 NovRoboticsComputer Vision and Pattern Recognition
The gist
Loop closure detection helps robots and devices recognize places they have visited before, which is important for accurate navigation. Usually, this process uses different separate steps that each need their own calculations, making it slower and less connected. The authors propose UniLoop, a single system that uses one model to handle all these steps together, sharing information and reducing redundant work. Their tests show that UniLoop detects repeated places more reliably and accurately than current methods.
Open → 2611.06378v1

Drone viewpoint planning ensures multiple quality images per area

Ground Sample Distance Constrained k-Coverage Viewpoint Planning via Set Multi-Cover Formulation

Abstract: Drone-based 3D reconstruction and inspection require viewpoints that satisfy image-quality constraints while observing each surface region multiple times, which overlap-based planning heuristics do not guarantee. We study ground sample distance (GSD)-constrained $k$-coverage viewpoint selection for structures represented by surface voxels. Candidate viewpoints are generated along outward surface directions, screened with distance, viewing-angle, field-of-view, and occlusion tests, and augmented for voxels with insufficient visibility. Selection is then posed as set multi-cover with per-voxel coverage demands. We propose a dynamically weighted greedy heuristic that prioritizes remaining coverage deficits and scarce viewing alternatives, and compare it with sequential overlap-based pruning and an integer linear programming (ILP) formulation on the same candidate pools. Across two structures, four voxel resolutions, and three coverage requirements, the proposed approach achieves full or near-full $k$-coverage with compact viewpoint sets, using only 3.00--16.39\% more viewpoints than ILP while offering an interpretable selection process based on coverage deficits and viewpoint availability.

Tue 27 OctRobotics
The gist
Getting good 3D models of buildings or structures with drones means taking multiple clear pictures of every spot. The authors show a way to choose drone viewpoints so that each part of the surface is seen enough times with good quality, using a new planning method. Their method smartly picks viewpoints by checking visibility and coverage needs, balancing quality and the number of drone positions. It works almost as well as a perfect but slow approach while being easier to understand and use.
Open → 2610.36450v1

Map auto-regressor generates vectorized roads and buildings

MARS - A foundational map auto-regressor

Abstract: Map generation tasks feature extensive non-structural vectorized data (e.g., points, polylines, and polygons) and thus pose significant challenges to common pixel-wise generative models. Conventional approaches use multiple stages, first segmenting these features at the pixel level and then performing vectorized post-processing, with errors and complexity compounding at each stage. Motivated by the recent success of auto-regressive language modeling, we propose the first map foundation model, named Map Auto-Regressor (MARS), that is capable of generating both multi-polyline road networks and polygon buildings in a unified manner. For training MARS, we collected to our knowledge the largest multi-class map extraction dataset totaling 3.4M examples, which we call MAP-3M. Across four road and building datasets, MARS outperforms or matches the performance of multistage baselines. Additionally, we develop a "Chat with MARS" capability that enables interactive human-in-the-loop map generation and correction, supported by the auto-regressive nature of our end-to-end approach. We release our MAP-3M dataset and project demo page at (1) https://huggingface.co/datasets/bag-lab/MAP-3M and (2) https://huggingface.co/spaces/bag-lab/MARS, respectively.

Wed 14 OctComputer Vision and Pattern Recognition
The gist
Map generation can be hard because maps contain lots of complex shapes like roads and buildings, which are tricky for usual image-based methods to draw correctly. The authors created MARS, a new model that can draw entire maps made of roads and buildings as connected lines and shapes, all in one step. They trained MARS on a very large dataset of maps they collected, and it works as well or better than older multi-step methods. MARS also allows people to interact with it to correct maps as they are created.
Open → 2610.20843v1

Geospatial vector geometry improves spatial reasoning in vision language models

Geospatial Vector Geometry as a Native Spatial Modality for Vision-Language Models

Abstract: Geospatial reasoning links VLMs to real-world action and decision making. Reliable reasoning requires accurate spatial information and effective spatial representations. However, current VLMs obtain this information mainly from language and vision. At larger geographic scales, limited visual observations make it difficult to establish accurate spatial references and recover precise spatial relations. This leads to a bottleneck for VLM spatial perception and reasoning. This paper proposes that geospatial vector geometry can become a native spatial modality within a VLM and introduces GeoVec-VLM. First, we develop a structured geometry representation that preserves the spatial structure and arrangement of vector objects, building an explicit spatial basis. Second, we introduce a geometry interaction strategy that brings this basis into multimodal computation, where it interacts with visual content and language references according to the queried spatial relations. Together, these designs allow perception and reasoning to operate on shared geometric references. Across five spatial reasoning tasks, GeoVec-VLM improves mean accuracy by about 11\% relative to the strongest baseline, maintains its gains under geographic transfer, and shows greater geometric consistency. Our results show that geospatial vector geometry offers a promising internal spatial basis for VLM spatial reasoning. https://github.com/TXFR/GeoVector-VLM

Mon 12 OctArtificial Intelligence
The gist
Understanding and reasoning about space in vision-language models is hard when looking at large geographic areas because they mostly rely on pictures and words. The authors found that using geospatial vector geometry as a built-in way to represent spatial information helps these models understand space better. They created a new model called GeoVec-VLM that uses this approach to connect geometry with images and language, improving accuracy by about 11 percent on spatial tasks. This method also works well when applied to new geographic areas and keeps spatial relationships more consistent.
Open → 2610.16769v1

Grus improve position tracking on curved spaces with new training methods

Manifold Navigation Learning Using Recurrent Networks: Efficient Training Techniques

Abstract: Position tracking is a well-studied problem in Euclidean navigation, yet comparatively little attention has been given to navigation on general manifolds, where motion is governed by the underlying geometry of the space. In this work, we formalize the problem of manifold navigation as a velocity-driven tracking task and propose a coordinate-chart-based learning framework in which learning models operate entirely in a Euclidean representation. Predicted trajectories are mapped to the manifold via an immersion function prior to error computation, allowing geometric consistency to be evaluated without explicitly enforcing manifold constraints during learning. Within this framework, we focus on Gated Recurrent Units (GRUs) for sequential position prediction on curved manifolds after comparisons with alternative recurrent architectures. We evaluate our GRU models, trained with pertinent custom loss functions, against both a standard MSE loss function and a naive vector-addition baseline. We evaluate our approach on both analytically (and numerically) defined manifolds, as well as a surface reconstructed from a LiDAR point cloud. Our results show that recurrent models can learn to approximate manifold-constrained dynamics from velocity sequences and consistently outperform Euclidean baselines. Our results demonstrate that certain custom loss functions significantly improve long-horizon prediction accuracy compared to conventional point-wise loss functions.

Fri 9 OctComputer Vision and Pattern Recognition
The gist
Tracking position is easier on flat surfaces but gets tricky on curved spaces called manifolds, which appear in many real-world situations. The authors introduce a way for computers to learn navigation on these curved spaces by using methods that work in flat coordinates but stay true to the curved shape. They tested a type of neural network called a Gated Recurrent Unit (GRU) and found it predicts movement on curved surfaces better than simpler methods. Also, they designed special training rules that help the model make better long-term predictions.
Open → 2610.14325v1

Spatial reasoning gap revealed in multimodal AI with GeoSpatialBench

Mind the Spatial Gap: A Benchmark for Geospatial Reasoning and Constrained Spatial Editing

Abstract: Understanding and manipulating spatial relations is fundamental to AI systems that reason about physical environments and construct grounded world models. However, existing evaluations of multimodal large language models (MLLMs) for spatial reasoning rely predominantly on static question-answering (QA) over small-scale or synthetic scenes, rarely probing whether models can modify spatial configurations while satisfying explicit geometric constraints. We introduce GeoSpatialBench, a benchmark designed to evaluate an escalating progression of spatial cognition, ranging from spatial relation understanding and geometric reasoning to constrained spatial editing. Grounded in real-world urban data across six geographically diverse cities, GeoSpatialBench evaluates a comprehensive taxonomy of 28 topological, directional, metric, and ordering relations across structured GeoJSON and map image modalities, with ground-truth references and constraints verified via GIS operations. Our empirical evaluation across frontier MLLMs uncovers a stark spatial gap: structured GeoJSON consistently outperforms map images, with multimodal combinations offering no reliable gain over GeoJSON alone, and while models show reasonable competence on qualitative topological and ordering, performance drops on quantitative metric relations. Crucially, editing performance on ordering relations declines relative to QA, and most failed edits violate at least one specified spatial constraint. These results highlight a gap between passively recognizing spatial relations and actively manipulating geometric configurations. GeoSpatialBench provides an open, rigorous diagnostic testbed to benchmark these spatial limitations, laying essential empirical groundwork for physically grounded world models and embodied spatial AI.

Fri 9 OctArtificial Intelligence
The gist
Many AI systems struggle to fully understand and change spatial relationships in maps and city layouts. The authors created GeoSpatialBench, a test that checks how well AI models handle real-world geographic data and maps with precise rules. They found AI models do better when using structured map data but often fail when asked to rearrange places while following strict rules. This shows that recognizing spatial layouts is easier for AI than actively editing them. Their test helps find and improve AI's weaknesses in understanding and handling spatial information.
Open → 2610.12881v1

Spatial indexes remain essential to accurately find places with large language models

Do We Still Need Gazetteers in the Era of LLMs? Chaining Retrieval with a Spatial Neuro-Symbolic Index

Abstract: Geographic information retrieval (GeoIR) tasks require systems to interpret ambiguous toponyms for downstream applications. Traditionally, toponym resolution relies on gazetteers to provide an explicit index of place entities and spatial relationships. Recently, gazetteer-free approaches seek to reduce dependence on handcrafted searches: dense retrieval utilizes text encoders to capture rich context, moving beyond the limitations of lexical search. However, text encoders implicitly assume that learned representations can function as reliable spatial-semantic indexes. In this paper, we evaluate this assumption through a spatial-semantic indexing setup: given a contextualized toponym mention, we retrieve the corresponding gazetteer entity represented by text derived from a gazetteer knowledge graph. We benchmark five frozen text encoders under two retrieval strategies: brute-force nearest-neighbor retrieval over entity representations, and a neuro-symbolic hierarchical beam search that constrains retrieval (i.e. chaining the search with gazetteer hierarchy). Experimental results reveal a distinct coarse-versus-fine trade-off. Unconstrained dense retrieval frequently incurs catastrophic spatial errors. Conversely, hierarchical constraints improve coarse geographic grounding, but still yield limited benefit for fine-grained localization metrics: vanilla text encoders fail to capture the fine-scale spatial fidelity encoded in gazetteers. Our code is publicly available at: https://doi.org/10.25439/rmt.31094269

Sun 4 OctInformation Retrieval
The gist
Finding the right place on a map can be tricky when place names mean different things in different contexts. The authors study whether big language models alone can replace traditional lists of place names called gazetteers, which also hold information about locations and relationships. They discover that while language models are good at roughly figuring out places, they often make big mistakes when trying to pinpoint exact locations without the structure gazetteers provide. Adding constraints from gazetteer hierarchies helps with broad location accuracy but doesn't fix all fine-scale errors. This shows that current language model methods still need gazetteers to reliably handle detailed spatial information.
Open → 2610.05028v1

3D Gaussian splatting improves remote sensing from few images

Remote Sensing Sparse-View 3D Gaussian Splatting via Depth Image-Based Rendering

Abstract: Remote sensing novel view synthesis under sparse observations remains challenging due to insufficient geometric constraints and limited cross-view supervision. Existing Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) methods are prone to overfitting and face challenges of depth ambiguities, missing cross-view information, and insufficient constraints in under-observed regions. To address these challenges, we propose DIBR-GS, a neural Gaussian Splatting framework that exploits Depth Image-Based Rendering (DIBR) to generate pseudo views for cross-view consistency supervision. Specifically, reliable geometric initialization is constructed by aligning monocular depth priors with sparse SfM reconstruction, and cross-view appearance priors are incorporated into neural Gaussian representations to enhance appearance modeling under sparse observations. Furthermore, we introduce a progressive DIBR-based pseudo-view supervision strategy to provide additional geometric and appearance constraints, enabling more complete reconstruction of weakly observed regions. In addition, a height-constrained anchor growth strategy is designed to suppress unreasonable Gaussian expansion. Experiments demonstrate that the proposed method achieves superior performance over existing approaches when training with only 3 input views. Compared with the previous best-performing method, it improves PSNR by 6.83 dB, with relative gains of 14\% in SSIM and 60\% in LPIPS, while maintaining competitive computational efficiency. Our code is available at https://github.com/kanehub/DIBR-GS

Mon 28 SeptComputer Vision and Pattern Recognition
The gist
Creating 3D views of a landscape or object from only a few photos is hard because there isn’t enough information to accurately guess the shapes and colors from all angles. The authors developed a new method that uses depth-based rendering to create extra virtual views, helping the system learn better 3D shapes and appearances. They also made sure the 3D parts don’t grow unrealistically tall, leading to more complete and accurate 3D models. Their method works well even when trained on just three images, outperforming previous techniques.
Open → 2609.35612v1

Super-resolution improves digital elevation models using satellite images

Guided Super-Resolution of Digital Elevation Models with Diffusion-Based Image Generators

Abstract: High-resolution digital surface models (DSMs) play an important role in urban analysis, 3D building reconstruction, and infrastructure monitoring, yet their availability remains limited due to the high cost and complexity of data acquisition. In contrast, coarse DSMs from commercial satellite missions are widely accessible, and high-resolution optical imagery is increasingly available from aerial and satellite platforms. We address the resulting mismatch in spatial resolution and propose a DSM superresolution approach that enhances 5 m DSMs to 0.5 m resolution, using guidance from high-resolution spectral images. Our method employs denoising diffusion to transfer information that is visible only in the image, like crisp outlines and detailed roof structures, into the elevation maps. In this way, surface details are reconstructed more accurately than with conventional interpolation or filtering techniques. Experiments on several cities in Central Europe demonstrate that the proposed approach produces high-quality DSMs with improved structural detail and accurate surface geometry. Our results highlight the potential of guided super-resolution with foundational image priors as a means of reconstructing high-resolution surface models.

Thu 10 SeptComputer Vision and Pattern Recognition
The gist
High-resolution 3D maps of surfaces, like cities, are very useful but hard and expensive to make. The authors found a way to improve low-resolution elevation maps by using clear satellite photos to add details. They use a special AI method called denoising diffusion to bring sharp shapes and features from the images into the elevation data. This method works better than older techniques and can create better 3D models of cities.
Open → 2609.11886v1