Recent advances in large language models (LLMs) have sparked growing interest in integrating language-driven techniques into trajectory prediction. By leveraging their semantic and reasoning capabilities, LLMs are reshaping how autonomous systems perceive, model, and predict trajectories. This survey provides a comprehensive overview of this emerging field, categorizing recent work into five directions: (1) Trajectory prediction via language modeling paradigms, (2) Direct trajectory prediction with pretrained language models, (3) Language-guided scene understanding for trajectory prediction, (4) Language-driven data generation for trajectory prediction, (5) Language-based reasoning and interpretability for trajectory prediction. For each, we analyze representative methods, highlight core design choices, and identify open challenges. This survey bridges natural language processing and trajectory prediction, offering a unified perspective on how language can enrich trajectory prediction.
Reference Audio-Visual Segmentation (Ref-AVS) aims to provide a pixel-wise scene understanding in Language-aided Audio-Visual Scenes (LAVS). This task requires the model to continuously segment objects referred to by text and audio from a video. Previous dual-modality methods always fail due to the lack of a third modality and the existing triple-modality method struggles with spatio-temporal consistency, leading to the target shift of different frames. In this work, we introduce a novel framework, termed SAM2-LOVE, which integrates textual, audio, and visual representations into a learnable token to prompt and align SAM2 for achieving Ref-AVS in the LAVS. Technically, our approach includes a multimodal fusion module aimed at improving multimodal understanding of SAM2, as well as token propagation and accumulation strategies designed to enhance spatio-temporal consistency without forgetting historical information. We conducted extensive experiments to demonstrate that SAM2-LOVE outperforms the SOTA by 8.5\% in $\mathcal{J\&F}$ on the Ref-AVS benchmark and showcase the simplicity and effectiveness of the components. Our code will be available here.
Remarkable progress in 2D Vision-Language Models (VLMs) has spurred interest in extending them to 3D settings for tasks like 3D Question Answering, Dense Captioning, and Visual Grounding. Unlike 2D VLMs that typically process images through an image encoder, 3D scenes, with their intricate spatial structures, allow for diverse model architectures. Based on their encoder design, this paper categorizes recent 3D VLMs into 3D object-centric, 2D image-based, and 3D scene-centric approaches. Despite the architectural similarity of 3D scene-centric VLMs to their 2D counterparts, they have exhibited comparatively lower performance compared with the latest 3D object-centric and 2D image-based approaches. To understand this gap, we conduct an in-depth analysis, revealing that 3D scene-centric VLMs show limited reliance on the 3D scene encoder, and the pre-train stage appears less effective than in 2D VLMs. Furthermore, we observe that data scaling benefits are less pronounced on larger datasets. Our investigation suggests that while these models possess cross-modal alignment capabilities, they tend to over-rely on linguistic cues and overfit to frequent answer distributions, thereby diminishing the effective utilization of the 3D encoder. To address these limitations and encourage genuine 3D scene understanding, we introduce a novel 3D Relevance Discrimination QA dataset designed to disrupt shortcut learning and improve 3D understanding. Our findings highlight the need for advanced evaluation and improved strategies for better 3D understanding in 3D VLMs.
Neural rendering has made significant strides in 3D reconstruction and novel view synthesis. With the integration with physics, it opens up new applications. The inverse problem of estimating physics from visual data, however, still remains challenging, limiting its effectiveness for applications like physically accurate digital twin creation in robotics and XR. Existing methods that incorporate physics into neural rendering frameworks typically require dense multi-view videos as input, making them impractical for scalable, real-world use. When presented with sparse multi-view videos, the sequential optimization strategy used by existing approaches introduces significant error accumulation, e.g., poor initial 3D reconstruction leads to bad material parameter estimation in subsequent stages. Instead of sequential optimization, directly optimizing all parameters at the same time also fails due to the highly non-convex and often non-differentiable nature of the problem. We propose ProJo4D, a progressive joint optimization framework that gradually increases the set of jointly optimized parameters guided by their sensitivity, leading to fully joint optimization over geometry, appearance, physical state, and material property. Evaluations on PAC-NeRF and Spring-Gaus datasets show that ProJo4D outperforms prior work in 4D future state prediction, novel view rendering of future state, and material parameter estimation, demonstrating its effectiveness in physically grounded 4D scene understanding. For demos, please visit the project webpage: https://daniel03c1.github.io/ProJo4D/
The rapid progress of navigation, manipulation, and vision models has made mobile manipulators capable in many specialized tasks. However, the open-world mobile manipulation (OWMM) task remains a challenge due to the need for generalization to open-ended instructions and environments, as well as the systematic complexity to integrate high-level decision making with low-level robot control based on both global scene understanding and current agent state. To address this complexity, we propose a novel multi-modal agent architecture that maintains multi-view scene frames and agent states for decision-making and controls the robot by function calling. A second challenge is the hallucination from domain shift. To enhance the agent performance, we further introduce an agentic data synthesis pipeline for the OWMM task to adapt the VLM model to our task domain with instruction fine-tuning. We highlight our fine-tuned OWMM-VLM as the first dedicated foundation model for mobile manipulators with global scene understanding, robot state tracking, and multi-modal action generation in a unified model. Through experiments, we demonstrate that our model achieves SOTA performance compared to other foundation models including GPT-4o and strong zero-shot generalization in real world. The project page is at https://github.com/HHYHRHY/OWMM-Agent
Music performances, characterized by dense and continuous audio as well as seamless audio-visual integration, present unique challenges for multimodal scene understanding and reasoning. Recent Music Performance Audio-Visual Question Answering (Music AVQA) datasets have been proposed to reflect these challenges, highlighting the continued need for more effective integration of audio-visual representations in complex question answering. However, existing Music AVQA methods often rely on dense and unoptimized representations, leading to inefficiencies in the isolation of key information, the reduction of redundancy, and the prioritization of critical samples. To address these challenges, we introduce Sparsify, a sparse learning framework specifically designed for Music AVQA. It integrates three sparsification strategies into an end-to-end pipeline and achieves state-of-the-art performance on the Music AVQA datasets. In addition, it reduces training time by 28.32% compared to its fully trained dense counterpart while maintaining accuracy, demonstrating clear efficiency gains. To further improve data efficiency, we propose a key-subset selection algorithm that selects and uses approximately 25% of MUSIC-AVQA v2.0 training data and retains 70-80% of full-data performance across models.
We introduce PhysGaia, a novel physics-aware dataset specifically designed for Dynamic Novel View Synthesis (DyNVS), encompassing both structured objects and unstructured physical phenomena. Unlike existing datasets that primarily focus on photorealistic reconstruction, PhysGaia is created to actively support physics-aware dynamic scene modeling. Our dataset provides complex dynamic scenarios with rich interactions among multiple objects, where they realistically collide with each other and exchange forces. Furthermore, it contains a diverse range of physical materials, such as liquid, gas, viscoelastic substance, and textile, which moves beyond the rigid bodies prevalent in existing datasets. All scenes in PhysGaia are faithfully generated to strictly adhere to physical laws, leveraging carefully selected material-specific physics solvers. To enable quantitative evaluation of physical modeling, our dataset provides essential ground-truth information, including 3D particle trajectories and physics parameters, e.g., viscosity. To facilitate research adoption, we also provide essential integration pipelines for using state-of-the-art DyNVS models with our dataset and report their results. By addressing the critical lack of datasets for physics-aware modeling, PhysGaia will significantly advance research in dynamic view synthesis, physics-based scene understanding, and deep learning models integrated with physical simulation -- ultimately enabling more faithful reconstruction and interpretation of complex dynamic scenes. Our datasets and codes are available in the project website, http://cvlab.snu.ac.kr/research/PhysGaia.
Recent advancements in 3D Gaussian Splatting (3D-GS) enable high-quality 3D scene reconstruction from RGB images. Many studies extend this paradigm for language-driven open-vocabulary scene understanding. However, most of them simply project 2D semantic features onto 3D Gaussians and overlook a fundamental gap between 2D and 3D understanding: a 3D object may exhibit various semantics from different viewpoints--a phenomenon we term view-dependent semantics. To address this challenge, we propose LaGa (Language Gaussians), which establishes cross-view semantic connections by decomposing the 3D scene into objects. Then, it constructs view-aggregated semantic representations by clustering semantic descriptors and reweighting them based on multi-view semantics. Extensive experiments demonstrate that LaGa effectively captures key information from view-dependent semantics, enabling a more comprehensive understanding of 3D scenes. Notably, under the same settings, LaGa achieves a significant improvement of +18.7% mIoU over the previous SOTA on the LERF-OVS dataset. Our code is available at: https://github.com/SJTU-DeepVisionLab/LaGa.
Cross-modal entity linking refers to the ability to align entities and their attributes across different modalities. While cross-modal entity linking is a fundamental skill needed for real-world applications such as multimodal code generation, fake news detection, or scene understanding, it has not been thoroughly studied in the literature. In this paper, we introduce a new task and benchmark to address this gap. Our benchmark, MATE, consists of 5.5k evaluation instances featuring visual scenes aligned with their textual representations. To evaluate cross-modal entity linking performance, we design a question-answering task that involves retrieving one attribute of an object in one modality based on a unique attribute of that object in another modality. We evaluate state-of-the-art Vision-Language Models (VLMs) and humans on this task, and find that VLMs struggle significantly compared to humans, particularly as the number of objects in the scene increases. Our analysis also shows that, while chain-of-thought prompting can improve VLM performance, models remain far from achieving human-level proficiency. These findings highlight the need for further research in cross-modal entity linking and show that MATE is a strong benchmark to support that progress.
Previous research has investigated the application of Multimodal Large Language Models (MLLMs) in understanding 3D scenes by interpreting them as videos. These approaches generally depend on comprehensive 3D data inputs, such as point clouds or reconstructed Bird's-Eye View (BEV) maps. In our research, we advance this field by enhancing the capability of MLLMs to understand and reason in 3D spaces directly from video data, without the need for additional 3D input. We propose a novel and efficient method, the Video-3D Geometry Large Language Model (VG LLM). Our approach employs a 3D visual geometry encoder that extracts 3D prior information from video sequences. This information is integrated with visual tokens and fed into the MLLM. Extensive experiments have shown that our method has achieved substantial improvements in various tasks related to 3D scene understanding and spatial reasoning, all directly learned from video sources. Impressively, our 4B model, which does not rely on explicit 3D data inputs, achieves competitive results compared to existing state-of-the-art methods, and even surpasses the Gemini-1.5-Pro in the VSI-Bench evaluations.
Generative models have demonstrated remarkable abilities in generating high-fidelity visual content. In this work, we explore how generative models can further be used not only to synthesize visual content but also to understand the properties of a scene given a natural image. We formulate scene understanding as an inverse generative modeling problem, where we seek to find conditional parameters of a visual generative model to best fit a given natural image. To enable this procedure to infer scene structure from images substantially different than those seen during training, we further propose to build this visual generative model compositionally from smaller models over pieces of a scene. We illustrate how this procedure enables us to infer the set of objects in a scene, enabling robust generalization to new test scenes with an increased number of objects of new shapes. We further illustrate how this enables us to infer global scene factors, likewise enabling robust generalization to new scenes. Finally, we illustrate how this approach can be directly applied to existing pretrained text-to-image generative models for zero-shot multi-object perception. Code and visualizations are at https://energy-based-model.github.io/compositional-inference.
Object orientation understanding represents a fundamental challenge in visual perception critical for applications like robotic manipulation and augmented reality. Current vision-language benchmarks fail to isolate this capability, often conflating it with positional relationships and general scene understanding. We introduce DORI (Discriminative Orientation Reasoning Intelligence), a comprehensive benchmark establishing object orientation perception as a primary evaluation target. DORI assesses four dimensions of orientation comprehension: frontal alignment, rotational transformations, relative directional relationships, and canonical orientation understanding. Through carefully curated tasks from 11 datasets spanning 67 object categories across synthetic and real-world scenarios, DORI provides insights on how multi-modal systems understand object orientations. Our evaluation of 15 state-of-the-art vision-language models reveals critical limitations: even the best models achieve only 54.2% accuracy on coarse tasks and 33.0% on granular orientation judgments, with performance deteriorating for tasks requiring reference frame shifts or compound rotations. These findings demonstrate the need for dedicated orientation representation mechanisms, as models show systematic inability to perform precise angular estimations, track orientation changes across viewpoints, and understand compound rotations - suggesting limitations in their internal 3D spatial representations. As the first diagnostic framework specifically designed for orientation awareness in multimodal systems, DORI offers implications for improving robotic control, 3D scene reconstruction, and human-AI interaction in physical environments. DORI data: https://huggingface.co/datasets/appledora/DORI-Benchmark
High-resolution (HR) remote sensing imagery plays a vital role in a wide range of applications, including urban planning and environmental monitoring. However, due to limitations in sensors and data transmission links, the images acquired in practice often suffer from resolution degradation. Remote Sensing Image Super-Resolution (RSISR) aims to reconstruct HR images from low-resolution (LR) inputs, providing a cost-effective and efficient alternative to direct HR image acquisition. Existing RSISR methods primarily focus on low-level characteristics in pixel space, while neglecting the high-level understanding of remote sensing scenes. This may lead to semantically inconsistent artifacts in the reconstructed results. Motivated by this observation, our work aims to explore the role of high-level semantic knowledge in improving RSISR performance. We propose a Semantic-Guided Super-Resolution framework, SeG-SR, which leverages Vision-Language Models (VLMs) to extract semantic knowledge from input images and uses it to guide the super resolution (SR) process. Specifically, we first design a Semantic Feature Extraction Module (SFEM) that utilizes a pretrained VLM to extract semantic knowledge from remote sensing images. Next, we propose a Semantic Localization Module (SLM), which derives a series of semantic guidance from the extracted semantic knowledge. Finally, we develop a Learnable Modulation Module (LMM) that uses semantic guidance to modulate the features extracted by the SR network, effectively incorporating high-level scene understanding into the SR pipeline. We validate the effectiveness and generalizability of SeG-SR through extensive experiments: SeG-SR achieves state-of-the-art performance on two datasets and consistently delivers performance improvements across various SR architectures. Codes can be found at https://github.com/Mr-Bamboo/SeG-SR.
A common starting point for drug design is to find small chemical groups or "fragments" that form interactions with distinct subregions in a protein binding pocket. The subsequent challenge is to assemble these fragments into a molecule that has high affinity to the protein, by adding chemical bonds between atoms in different fragments. This "molecule assembly" task is particularly challenging because, initially, fragment positions are known only approximately. Prior methods for spatial graph completion-adding missing edges to a graph whose nodes have associated spatial coordinates-either treat node positions as fixed or adjust node positions before predicting edges. The fact that these methods treat geometry and topology prediction separately limits their ability to reconcile noisy geometries and plausible connectivities. To address this limitation, we introduce EdGr, a spatial graph diffusion model that reasons jointly over geometry and topology of molecules to simultaneously predict fragment positions and inter-fragment bonds. Importantly, predicted edge likelihoods directly influence node position updates during the diffusion denoising process, allowing connectivity cues to guide spatial movements, and vice versa. EdGr substantially outperforms previous methods on the molecule assembly task and maintains robust performance as noise levels increase. Beyond drug discovery, our approach of explicitly coupling geometry and topology prediction is broadly applicable to spatial graph completion problems, such as neural circuit reconstruction, 3D scene understanding, and sensor network design.
3D semantic occupancy prediction offers an intuitive and efficient scene understanding and has attracted significant interest in autonomous driving perception. Existing approaches either rely on full supervision, which demands costly voxel-level annotations, or on self-supervision, which provides limited guidance and yields suboptimal performance. To address these challenges, we propose OccLE, a Label-Efficient 3D Semantic Occupancy Prediction that takes images and LiDAR as inputs and maintains high performance with limited voxel annotations. Our intuition is to decouple the semantic and geometric learning tasks and then fuse the learned feature grids from both tasks for the final semantic occupancy prediction. Therefore, the semantic branch distills 2D foundation model to provide aligned pseudo labels for 2D and 3D semantic learning. The geometric branch integrates image and LiDAR inputs in cross-plane synergy based on their inherency, employing semi-supervision to enhance geometry learning. We fuse semantic-geometric feature grids through Dual Mamba and incorporate a scatter-accumulated projection to supervise unannotated prediction with aligned pseudo labels. Experiments show that OccLE achieves competitive performance with only 10% of voxel annotations, reaching a mIoU of 16.59% on the SemanticKITTI validation set.
We propose a novel framework for comprehensive indoor 3D reconstruction using Gaussian representations, called OmniIndoor3D. This framework enables accurate appearance, geometry, and panoptic reconstruction of diverse indoor scenes captured by a consumer-level RGB-D camera. Since 3DGS is primarily optimized for photorealistic rendering, it lacks the precise geometry critical for high-quality panoptic reconstruction. Therefore, OmniIndoor3D first combines multiple RGB-D images to create a coarse 3D reconstruction, which is then used to initialize the 3D Gaussians and guide the 3DGS training. To decouple the optimization conflict between appearance and geometry, we introduce a lightweight MLP that adjusts the geometric properties of 3D Gaussians. The introduced lightweight MLP serves as a low-pass filter for geometry reconstruction and significantly reduces noise in indoor scenes. To improve the distribution of Gaussian primitives, we propose a densification strategy guided by panoptic priors to encourage smoothness on planar surfaces. Through the joint optimization of appearance, geometry, and panoptic reconstruction, OmniIndoor3D provides comprehensive 3D indoor scene understanding, which facilitates accurate and robust robotic navigation. We perform thorough evaluations across multiple datasets, and OmniIndoor3D achieves state-of-the-art results in appearance, geometry, and panoptic reconstruction. We believe our work bridges a critical gap in indoor 3D reconstruction. The code will be released at: https://ucwxb.github.io/OmniIndoor3D/
In this study, we present a novel LiDAR-based semantic segmentation framework tailored for autonomous forklifts operating in complex outdoor environments. Central to our approach is the integration of a dual LiDAR system, which combines forward-facing and downward-angled LiDAR sensors to enable comprehensive scene understanding, specifically tailored for industrial material handling tasks. The dual configuration improves the detection and segmentation of dynamic and static obstacles with high spatial precision. Using high-resolution 3D point clouds captured from two sensors, our method employs a lightweight yet robust approach that segments the point clouds into safety-critical instance classes such as pedestrians, vehicles, and forklifts, as well as environmental classes such as driveable ground, lanes, and buildings. Experimental validation demonstrates that our approach achieves high segmentation accuracy while satisfying strict runtime requirements, establishing its viability for safety-aware, fully autonomous forklift navigation in dynamic warehouse and yard environments.
Adaptive navigation in unfamiliar environments is crucial for household service robots but remains challenging due to the need for both low-level path planning and high-level scene understanding. While recent vision-language model (VLM) based zero-shot approaches reduce dependence on prior maps and scene-specific training data, they face significant limitations: spatiotemporal discontinuity from discrete observations, unstructured memory representations, and insufficient task understanding leading to navigation failures. We propose DORAEMON (Decentralized Ontology-aware Reliable Agent with Enhanced Memory Oriented Navigation), a novel cognitive-inspired framework consisting of Ventral and Dorsal Streams that mimics human navigation capabilities. The Dorsal Stream implements the Hierarchical Semantic-Spatial Fusion and Topology Map to handle spatiotemporal discontinuities, while the Ventral Stream combines RAG-VLM and Policy-VLM to improve decision-making. Our approach also develops Nav-Ensurance to ensure navigation safety and efficiency. We evaluate DORAEMON on the HM3D, MP3D, and GOAT datasets, where it achieves state-of-the-art performance on both success rate (SR) and success weighted by path length (SPL) metrics, significantly outperforming existing methods. We also introduce a new evaluation metric (AORI) to assess navigation intelligence better. Comprehensive experiments demonstrate DORAEMON's effectiveness in zero-shot autonomous navigation without requiring prior map building or pre-training.
The segmentation of kidney stones is regarded as a critical preliminary step to enable the identification of urinary stone types through machine- or deep-learning-based approaches. In urology, manual segmentation is considered tedious and impractical due to the typically large scale of image databases and the continuous generation of new data. In this study, the potential of the Segment Anything Model (SAM) -- a state-of-the-art deep learning framework -- is investigated for the automation of kidney stone segmentation. The performance of SAM is evaluated in comparison to traditional models, including U-Net, Residual U-Net, and Attention U-Net, which, despite their efficiency, frequently exhibit limitations in generalizing to unseen datasets. The findings highlight SAM's superior adaptability and efficiency. While SAM achieves comparable performance to U-Net on in-distribution data (Accuracy: 97.68 + 3.04; Dice: 97.78 + 2.47; IoU: 95.76 + 4.18), it demonstrates significantly enhanced generalization capabilities on out-of-distribution data, surpassing all U-Net variants by margins of up to 23 percent.
Underwater images are often affected by complex degradations such as light absorption, scattering, color casts, and artifacts, making enhancement critical for effective object detection, recognition, and scene understanding in aquatic environments. Existing methods, especially diffusion-based approaches, typically rely on synthetic paired datasets due to the scarcity of real underwater references, introducing bias and limiting generalization. Furthermore, fine-tuning these models can degrade learned priors, resulting in unrealistic enhancements due to domain shifts. To address these challenges, we propose UDAN-CLIP, an image-to-image diffusion framework pre-trained on synthetic underwater datasets and enhanced with a customized classifier based on vision-language model, a spatial attention module, and a novel CLIP-Diffusion loss. The classifier preserves natural in-air priors and semantically guides the diffusion process, while the spatial attention module focuses on correcting localized degradations such as haze and low contrast. The proposed CLIP-Diffusion loss further strengthens visual-textual alignment and helps maintain semantic consistency during enhancement. The proposed contributions empower our UDAN-CLIP model to perform more effective underwater image enhancement, producing results that are not only visually compelling but also more realistic and detail-preserving. These improvements are consistently validated through both quantitative metrics and qualitative visual comparisons, demonstrating the model's ability to correct distortions and restore natural appearance in challenging underwater conditions.