Jul 2026· Signal Processing and Communications Applications Conference· pp. 1-4· 0 citations· 9 references
Abstract
In emergency scenarios such as fires or earthquakes, rapid and accurate situational awareness of the scene is critical for decision-making processes. Operators may inaccurately visualize descriptions received from victims under stress. In this study, a web-based system is proposed that instantly converts natural language environment descriptions into three-dimensional (3D) scene visualizations. Although Generative AI approaches produce photorealistic images, they carry the risk of hallucination. Therefore, this study adopts a deterministic Scene Assembly approach that prioritizes spatial consistency. The system converts user text into a structured JSON format via an LLM-based parser, validates physical consistency through a rulebased spatial inference layer, and computes evacuation routes using the A* algorithm. Experimental results on 100 scenarios show that while LLM-only achieves 76.3% overall accuracy, the addition of the spatial constraint layer raises this to 85.7%.
In the context of urban planning, architects are normally instructed with creating presentation images that visualize proposed buildings within their urban context. This work aims to develop a GenAI model for automatically generating architectural presentation images in urban scenes, with emphasis on model optimization. To achieve this, we developed Mask-based Weighted Conditional Flow Matching (MWCFM), which extends Flow Matching by introducing contextual masks for precise feature focusing. This enables targeted training on critical spatial elements relevant to urban planning. Our trained model learns from urban street-view data while adhering to specific style-guidelines, which are integrated into training through the loss function. Furthermore, the model's performance is evaluated using application related metrics, derived from presentation image style guidelines.
Katharina Roth, Eva Hagen, Alexander Bartscher et al.· 0 citations
Beyond perception, reasoning is essential in remote sensing for advanced interpretation, inference, and decision-making. Recent advances in large language models (LLMs) have enabled tool-augmented agents that leverage external tools to perform complex analytical tasks. However, existing studies in remote sensing primarily focus on perception-oriented tasks, leaving cognitive geospatial reasoning largely underexplored. To address this gap, we introduce TerraLogic, a benchmark for geospatial reasoning. TerraLogic comprises 545 scenario-driven, hierarchy-aware tasks, such as hazard vulnerability assessment, urban heat island analysis, and forest fragmentation dynamics, spanning optical, Synthetic Aperture Radar (SAR), and infrared (IR) imagery. It advances evaluation beyond recognition and monitoring toward cognitive-level geospatial analysis. To facilitate evaluation on TerraLogic, we further propose HieraPlan, a tool-augmented agent that organizes toolkits into functional hierarchies and performs fault-tolerant reasoning. HieraPlan enables structured abstraction, robust recovery from tool failures, and stable long-horizon planning. Extensive experiments demonstrate that current approaches struggle with hierarchical geospatial reasoning, while HieraPlan provides a strong baseline with improved reasoning, cross-modal generalization, and error handling. The dataset and agent code are publicly available at https://github.com/Ireliya/TerraLogic.
Yuhang Yan, Linchao Mou, Bokang Yang et al.· 1 citation
The Cinematic Pre-Visualization System is an end-to-end pipeline that converts natural language scene descriptions into structured three-dimensional (3D) previsualization scenes in Blender and Unity. The system employs Natural Language Processing (NLP) techniques using the spaCy library to extract semantic scene elements including characters, objects, spatial relationships, lighting conditions, and camera directives from textual input. Extracted information is serialized into a platform-agnostic JSON intermediate representation, which drives automated 3D scene construction in Blender via the bpy Python API and in Unity via C# scripting. Experimental evaluation on twenty diverse cinematic scene descriptions demonstrates an average entity extraction accuracy of 87.5%, spatial relationship accuracy of 82.3%, lighting keyword detection of 94.1%, and camera directive extraction of 90.0%. The total pipeline latency from text input to rendered scene is under 6 seconds, validating the system's suitability for iterative use in real-world pre-production workflows. The system significantly reduces scene setup time and lowers the technical barrier for previsualization, offering a practical tool for the film, animation, and game production industries.
B. Dheepa, Monish J, Praveen R et al.· Advanced International Journ...· 0 citations
Urban street retrofitting is increasingly used to improve greenery, walkability, and perceived safety, yet planners and communities often lack intuitive visualizations of how a street may look after such interventions. This paper presents a perception-aware framework for green streetscape redesign that transforms a real street-view image and a high-level design goal into a realistic visualization of a retrofitted street. The proposed framework integrates an MLLM-guided planner for structured redesign operations, a rule-based compiler for semantic mask editing, a ControlNet-guided diffusion renderer for candidate generation, and a perception-aware selector for choosing the final design. Experiments on Cityscapes and Mapillary Vistas, show that the method achieves a favorable balance among environmental improvement, image realism, and structural preservation. Additional comparisons further demonstrate the value of multimodal planning and perception-aware ranking. These results suggest that controllable generative models can provide practical support for urban street retrofit visualization and assessment.
Hongkun Wang, Fei Li· Digital Signal and Computer...· 0 citations
Edge-deployed systems such as autonomous robots, AR/XR devices, and emergency-response handhelds require accurate indoor topological representations, yet existing sensor-based and expert-curated mapping methods are impractical for crowd-sourced, resource-constrained deployment. Additionally, current large language model (LLM) approaches to indoor topology inference lack a systematic framework for evaluating the factors that govern performance. We present Lang2Graph, an experimental framework for indoor topological graph inference from natural-language navigational instructions that isolates four governing factors: instruction structure, metadata clarity, prompting strategy, and model size and reasoning capability. We propose the Independent Prompt Executor (IPE), a prompting strategy that decomposes graph construction into independent per-instruction reasoning steps, preventing error propagation. To support factor-level evaluation, we develop a fully synthetic dataset and an augmented benchmark (R2R-AUG) covering idealized and real-world conditions. Our evaluation across multiple model families show that structured instructions, clear metadata, and IPE improve precision, recall, and F1 by 44%, 45%, and 47%, respectively. Reasoning-aligned open-source models of moderate scale (14B parameters) outperform larger proprietary models on the most challenging instruction categories, indicating that reliable indoor topology inference is achievable without cloud-scale dependencies and establishes a viable path toward on-device edge deployment.
Moamin Ibrahim, Yaqoob Ansari, Khaled A. Harras et al.· International Conference on...· 0 citations
Multimodal large language models (MLLMs) can interpret a street view, but urban agency depends on whether such local evidence remains useful after the agent starts to move. In this paper, we investigate how far current MLLM agents can turn local urban perception into reliable action in a complicated real-scale city. We propose UrbanGround, the first sandbox to make this question testable in a physically constrained replica of Hong Kong built from territory-wide 3D geospatial data. UrbanGround supports closed-loop interaction from a first-person view and provides an interactive map for navigation. Agents can directly enter the 3D city and explore from a first-person view. Our analysis follows the growth of the spatial problem through three research questions. We first test whether an agent can ground a local scene well enough to answer spatial questions after active observation. Then we ask whether that grounding supports navigation as destinations become farther away and less explicit. Finally, we examine whether the resulting behavior survives changes in route availability and pedestrian motion. Contemporary MLLM agents usually show useful atomic abilities in visual recognition and short-range spatial reasoning, while orientation and pedestrian-aware movement remain unreliable. Their central failure emerges over extended exploration, where local abilities do not compose into sustained goal-directed behavior and errors accumulate without effective correction. We hope UrbanGround will support broader study of how far current MLLM agents can explore reliably in complex, open-ended urban environments.
Tianjie Ju, Zheng Wu, Yueqing Sun et al.· 0 citations