Scene flow provides crucial motion information for autonomous driving. Recent LiDAR scene flow models utilize the rigid-motion assumption at the instance level, assuming objects are rigid bodies. However, these instance-level methods are not suitable for sparse radar point clouds. In this work, we present a novel $\textbf{T}$raffic-$\textbf{A}$ware $\textbf{R}$adar $\textbf{S}$cene flow estimation method, named $\textbf{TARS}$, which utilizes the motion rigidity at the traffic level. To address the challenges in radar scene flow, we perform object detection and scene flow jointly and boost the latter. We incorporate the feature map from the object detector, trained with detection losses, to make radar scene flow aware of the environment and road users. Therefrom, we construct a Traffic Vector Field (TVF) in the feature space, enabling a holistic traffic-level scene understanding in our scene flow branch. When estimating the scene flow, we consider both point-level motion cues from point neighbors and traffic-level consistency of rigid motion within the space. TARS outperforms the state of the art on a proprietary dataset and the View-of-Delft dataset, improving the benchmarks by 23% and 15%, respectively.
3D semantic occupancy has rapidly become a research focus in the fields of robotics and autonomous driving environment perception due to its ability to provide more realistic geometric perception and its closer integration with downstream tasks. By performing occupancy prediction of the 3D space in the environment, the ability and robustness of scene understanding can be effectively improved. However, existing occupancy prediction tasks are primarily modeled using voxel or point cloud-based approaches: voxel-based network structures often suffer from the loss of spatial information due to the voxelization process, while point cloud-based methods, although better at retaining spatial location information, face limitations in representing volumetric structural details. To address this issue, we propose a dual-modal prediction method based on 3D Gaussian sets and sparse points, which balances both spatial location and volumetric structural information, achieving higher accuracy in semantic occupancy prediction. Specifically, our method adopts a Transformer-based architecture, taking 3D Gaussian sets, sparse points, and queries as inputs. Through the multi-layer structure of the Transformer, the enhanced queries and 3D Gaussian sets jointly contribute to the semantic occupancy prediction, and an adaptive fusion mechanism integrates the semantic outputs of both modalities to generate the final prediction results. Additionally, to further improve accuracy, we dynamically refine the point cloud at each layer, allowing for more precise location information during occupancy prediction. We conducted experiments on the Occ3DnuScenes dataset, and the experimental results demonstrate superior performance of the proposed method on IoU based metrics.
Object-centric learning is fundamental to human vision and crucial for models requiring complex reasoning. Traditional approaches rely on slot-based bottlenecks to learn object properties explicitly, while recent self-supervised vision models like DINO have shown emergent object understanding. However, DINO representations primarily capture global scene features, often confounding individual object attributes. We investigate the effectiveness of DINO representations and slot-based methods for multi-object instance retrieval. Our findings reveal that DINO representations excel at capturing global object attributes such as object shape and size, but struggle with object-level details like colour, whereas slot-based representations struggle at both global and object-level understanding. To address this, we propose a method that combines global and local features by augmenting DINO representations with object-centric latent vectors from a Variational Autoencoder trained on segmented image patches that are extracted from the DINO features. This approach improves multi-object instance retrieval performance, bridging the gap between global scene understanding and fine-grained object representation without requiring full model retraining.
Comprehensive and consistent dynamic scene understanding from camera input is essential for advanced autonomous systems. Traditional camera-based perception tasks like 3D object tracking and semantic occupancy prediction lack either spatial comprehensiveness or temporal consistency. In this work, we introduce a brand-new task, Camera-based 4D Panoptic Occupancy Tracking, which simultaneously addresses panoptic occupancy segmentation and object tracking from camera-only input. Furthermore, we propose TrackOcc, a cutting-edge approach that processes image inputs in a streaming, end-to-end manner with 4D panoptic queries to address the proposed task. Leveraging the localization-aware loss, TrackOcc enhances the accuracy of 4D panoptic occupancy tracking without bells and whistles. Experimental results demonstrate that our method achieves state-of-the-art performance on the Waymo dataset. The source code will be released at https://github.com/Tsinghua-MARS-Lab/TrackOcc.
The emergence of large language models (LLMs) opens new frontiers for unmanned aerial vehicle (UAVs), yet existing systems remain confined to predefined tasks due to hardware-software co-design challenges. This paper presents the first aerial intelligent agent capable of open-world task execution through tight integration of LLM-based reasoning and robotic autonomy. Our hardware-software co-designed system addresses two fundamental limitations: (1) Onboard LLM operation via an edge-optimized computing platform, achieving 5-6 tokens/sec inference for 14B-parameter models at 220W peak power; (2) A bidirectional cognitive architecture that synergizes slow deliberative planning (LLM task planning) with fast reactive control (state estimation, mapping, obstacle avoidance, and motion planning). Validated through preliminary results using our prototype, the system demonstrates reliable task planning and scene understanding in communication-constrained environments, such as sugarcane monitoring, power grid inspection, mine tunnel exploration, and biological observation applications. This work establishes a novel framework for embodied aerial artificial intelligence, bridging the gap between task planning and robotic autonomy in open environments.
Until recently, robotics safety research was predominantly about collision avoidance and hazard reduction in the immediate vicinity of a robot. Since the advent of large vision and language models (VLMs), robots are now also capable of higher-level semantic scene understanding and natural language interactions with humans. Despite their known vulnerabilities (e.g. hallucinations or jail-breaking), VLMs are being handed control of robots capable of physical contact with the real world. This can lead to dangerous behaviors, making semantic safety for robots a matter of immediate concern. Our contributions in this paper are two fold: first, to address these emerging risks, we release the ASIMOV Benchmark, a large-scale and comprehensive collection of datasets for evaluating and improving semantic safety of foundation models serving as robot brains. Our data generation recipe is highly scalable: by leveraging text and image generation techniques, we generate undesirable situations from real-world visual scenes and human injury reports from hospitals. Secondly, we develop a framework to automatically generate robot constitutions from real-world data to steer a robot's behavior using Constitutional AI mechanisms. We propose a novel auto-amending process that is able to introduce nuances in written rules of behavior; this can lead to increased alignment with human preferences on behavior desirability and safety. We explore trade-offs between generality and specificity across a diverse set of constitutions of different lengths, and demonstrate that a robot is able to effectively reject unconstitutional actions. We measure a top alignment rate of 84.3% on the ASIMOV Benchmark using generated constitutions, outperforming no-constitution baselines and human-written constitutions. Data is available at asimov-benchmark.github.io
Background: We evaluate SAM 2 for surgical scene understanding by examining its semantic segmentation capabilities for organs/tissues both in zero-shot scenarios and after fine-tuning. Methods: We utilized five public datasets to evaluate and fine-tune SAM 2 for segmenting anatomical tissues in surgical videos/images. Fine-tuning was applied to the image encoder and mask decoder. We limited training subsets from 50 to 400 samples per class to better model real-world constraints with data acquisition. The impact of dataset size on fine-tuning performance was evaluated with weighted mean Dice coefficient (WMDC), and the results were also compared against previously reported state-of-the-art (SOTA) results. Results: SurgiSAM 2, a fine-tuned SAM 2 model, demonstrated significant improvements in segmentation performance, achieving a 17.9% relative WMDC gain compared to the baseline SAM 2. Increasing prompt points from 1 to 10 and training data scale from 50/class to 400/class enhanced performance; the best WMDC of 0.92 on the validation subset was achieved with 10 prompt points and 400 samples per class. On the test subset, this model outperformed prior SOTA methods in 24/30 (80%) of the classes with a WMDC of 0.91 using 10-point prompts. Notably, SurgiSAM 2 generalized effectively to unseen organ classes, achieving SOTA on 7/9 (77.8%) of them. Conclusion: SAM 2 achieves remarkable zero-shot and fine-tuned performance for surgical scene segmentation, surpassing prior SOTA models across several organ classes of diverse datasets. This suggests immense potential for enabling automated/semi-automated annotation pipelines, thereby decreasing the burden of annotations facilitating several surgical applications.
Lane topology extraction involves detecting lanes and traffic elements and determining their relationships, a key perception task for mapless autonomous driving. This task requires complex reasoning, such as determining whether it is possible to turn left into a specific lane. To address this challenge, we introduce neuro-symbolic methods powered by vision-language foundation models (VLMs). Existing approaches have notable limitations: (1) Dense visual prompting with VLMs can achieve strong performance but is costly in terms of both financial resources and carbon footprint, making it impractical for robotics applications. (2) Neuro-symbolic reasoning methods for 3D scene understanding fail to integrate visual inputs when synthesizing programs, making them ineffective in handling complex corner cases. To this end, we propose a fast-slow neuro-symbolic lane topology extraction algorithm, named Chameleon, which alternates between a fast system that directly reasons over detected instances using synthesized programs and a slow system that utilizes a VLM with a chain-of-thought design to handle corner cases. Chameleon leverages the strengths of both approaches, providing an affordable solution while maintaining high performance. We evaluate the method on the OpenLane-V2 dataset, showing consistent improvements across various baseline detectors. Our code, data, and models are publicly available at https://github.com/XR-Lee/neural-symbolic
Large multimodal models (LMMs) excel in scene understanding but struggle with fine-grained spatiotemporal reasoning due to weak alignment between linguistic and visual representations. Existing methods map textual positions and durations into the visual space encoded from frame-based videos, but suffer from temporal sparsity that limits language-vision temporal coordination. To address this issue, we introduce LLaFEA (Large Language and Frame-Event Assistant) to leverage event cameras for temporally dense perception and frame-event fusion. Our approach employs a cross-attention mechanism to integrate complementary spatial and temporal features, followed by self-attention matching for global spatio-temporal associations. We further embed textual position and duration tokens into the fused visual space to enhance fine-grained alignment. This unified framework ensures robust spatio-temporal coordinate alignment, enabling LMMs to interpret scenes at any position and any time. In addition, we construct a dataset of real-world frames-events with coordinate instructions and conduct extensive experiments to validate the effectiveness of the proposed method.
Language-guided 3D scene understanding is important for advancing applications in robotics, AR/VR, and human-computer interaction, enabling models to comprehend and interact with 3D environments through natural language. While 2D vision-language models (VLMs) have achieved remarkable success in 2D VQA tasks, progress in the 3D domain has been significantly slower due to the complexity of 3D data and the high cost of manual annotations. In this work, we introduce SplatTalk, a novel method that uses a generalizable 3D Gaussian Splatting (3DGS) framework to produce 3D tokens suitable for direct input into a pretrained LLM, enabling effective zero-shot 3D visual question answering (3D VQA) for scenes with only posed images. During experiments on multiple benchmarks, our approach outperforms both 3D models trained specifically for the task and previous 2D-LMM-based models utilizing only images (our setting), while achieving competitive performance with state-of-the-art 3D LMMs that additionally utilize 3D inputs.
Amodal instance segmentation, which aims to detect and segment both visible and invisible parts of objects in images, plays a crucial role in various applications including autonomous driving, robotic manipulation, and scene understanding. While existing methods require training both front-end detectors and mask decoders jointly, this approach lacks flexibility and fails to leverage the strengths of pre-existing modal detectors. To address this limitation, we propose SAMEO, a novel framework that adapts the Segment Anything Model (SAM) as a versatile mask decoder capable of interfacing with various front-end detectors to enable mask prediction even for partially occluded objects. Acknowledging the constraints of limited amodal segmentation datasets, we introduce Amodal-LVIS, a large-scale synthetic dataset comprising 300K images derived from the modal LVIS and LVVIS datasets. This dataset significantly expands the training data available for amodal segmentation research. Our experimental results demonstrate that our approach, when trained on the newly extended dataset, including Amodal-LVIS, achieves remarkable zero-shot performance on both COCOA-cls and D2SA benchmarks, highlighting its potential for generalization to unseen scenarios.
Camera-based 3D semantic scene completion (SSC) provides dense geometric and semantic perception for autonomous driving. However, images provide limited information making the model susceptible to geometric ambiguity caused by occlusion and perspective distortion. Existing methods often lack explicit semantic modeling between objects, limiting their perception of 3D semantic context. To address these challenges, we propose a novel method VLScene: Vision-Language Guidance Distillation for Camera-based 3D Semantic Scene Completion. The key insight is to use the vision-language model to introduce high-level semantic priors to provide the object spatial context required for 3D scene understanding. Specifically, we design a vision-language guidance distillation process to enhance image features, which can effectively capture semantic knowledge from the surrounding environment and improve spatial context reasoning. In addition, we introduce a geometric-semantic sparse awareness mechanism to propagate geometric structures in the neighborhood and enhance semantic information through contextual sparse interactions. Experimental results demonstrate that VLScene achieves rank-1st performance on challenging benchmarks--SemanticKITTI and SSCBench-KITTI-360, yielding remarkably mIoU scores of 17.52 and 19.10, respectively.
Minimally invasive surgery (MIS) has transformed clinical practice by reducing recovery times, minimizing complications, and enhancing precision. Nonetheless, MIS inherently relies on indirect visualization and precise instrument control, posing unique challenges. Recent advances in artificial intelligence have enabled real-time surgical scene understanding through techniques such as image classification, object detection, and segmentation, with scene reconstruction emerging as a key element for enhanced intraoperative guidance. Although neural radiance fields (NeRFs) have been explored for this purpose, their substantial data requirements and slow rendering inhibit real-time performance. In contrast, 3D Gaussian Splatting (3DGS) offers a more efficient alternative, achieving state-of-the-art performance in dynamic surgical scene reconstruction. In this work, we introduce Feature-EndoGaussian (FEG), an extension of 3DGS that integrates 2D segmentation cues into 3D rendering to enable real-time semantic and scene reconstruction. By leveraging pretrained segmentation foundation models, FEG incorporates semantic feature distillation within the Gaussian deformation framework, thereby enhancing both reconstruction fidelity and segmentation accuracy. On the EndoNeRF dataset, FEG achieves superior performance (SSIM of 0.97, PSNR of 39.08, and LPIPS of 0.03) compared to leading methods. Additionally, on the EndoVis18 dataset, FEG demonstrates competitive class-wise segmentation metrics while balancing model size and real-time performance.
Accurate motion forecasting is crucial for safe autonomous driving (AD). This study proposes CoT-Drive, a novel approach that enhances motion forecasting by leveraging large language models (LLMs) and a chain-of-thought (CoT) prompting method. We introduce a teacher-student knowledge distillation strategy to effectively transfer LLMs' advanced scene understanding capabilities to lightweight language models (LMs), ensuring that CoT-Drive operates in real-time on edge devices while maintaining comprehensive scene understanding and generalization capabilities. By leveraging CoT prompting techniques for LLMs without additional training, CoT-Drive generates semantic annotations that significantly improve the understanding of complex traffic environments, thereby boosting the accuracy and robustness of predictions. Additionally, we present two new scene description datasets, Highway-Text and Urban-Text, designed for fine-tuning lightweight LMs to generate context-specific semantic annotations. Comprehensive evaluations of five real-world datasets demonstrate that CoT-Drive outperforms existing models, highlighting its effectiveness and efficiency in handling complex traffic scenarios. Overall, this study is the first to consider the practical application of LLMs in this field. It pioneers the training and use of a lightweight LLM surrogate for motion forecasting, setting a new benchmark and showcasing the potential of integrating LLMs into AD systems.
Pre-trained vision models (PVMs) are fundamental to modern robotics, yet their optimal configuration remains unclear. Through systematic evaluation, we find that while DINO and iBOT outperform MAE across visuomotor control and perception tasks, they struggle when trained on non-(single-)object-centric (NOC) data--a limitation strongly correlated with their diminished ability to learn object-centric representations. This investigation indicates that the ability to form object-centric representations from the non-object-centric robotics dataset is the key to success for PVMs. Motivated by this discovery, we designed SlotMIM, a method that induces object-centric representations by introducing a semantic bottleneck to reduce the number of prototypes to encourage the emergence of objectness as well as cross-view consistency regularization for encouraging multiview invariance. Our experiments encompass pre-training on object-centric, scene-centric, web-crawled, and ego-centric data. Across all settings, our approach learns transferrable representations and achieves significant improvements over prior work in image recognition, scene understanding, and robot learning evaluations. When scaled up with million-scale datasets, our method also demonstrates superior data efficiency and scalability. Our code and models are publicly available at https://github.com/CVMI-Lab/SlotMIM.
Depth ambiguity is a fundamental challenge in spatial scene understanding, especially in transparent scenes where single-depth estimates fail to capture full 3D structure. Existing models, limited to deterministic predictions, overlook real-world multi-layer depth. To address this, we introduce a paradigm shift from single-prediction to multi-hypothesis spatial foundation models. We first present \texttt{MD-3k}, a benchmark exposing depth biases in expert and foundational models through multi-layer spatial relationship labels and new metrics. To resolve depth ambiguity, we propose Laplacian Visual Prompting (LVP), a training-free spectral prompting technique that extracts hidden depth from pre-trained models via Laplacian-transformed RGB inputs. By integrating LVP-inferred depth with standard RGB-based estimates, our approach elicits multi-layer depth without model retraining. Extensive experiments validate the effectiveness of LVP in zero-shot multi-layer depth estimation, unlocking more robust and comprehensive geometry-conditioned visual generation, 3D-grounded spatial reasoning, and temporally consistent video-level depth inference. Our benchmark and code will be available at https://github.com/Xiaohao-Xu/Ambiguity-in-Space.
For scene understanding in unstructured environments, an accurate and uncertainty-aware metric-semantic mapping is required to enable informed action selection by autonomous systems. Existing mapping methods often suffer from overconfident semantic predictions, and sparse and noisy depth sensing, leading to inconsistent map representations. In this paper, we therefore introduce EvidMTL, a multi-task learning framework that uses evidential heads for depth estimation and semantic segmentation, enabling uncertainty-aware inference from monocular RGB images. To enable uncertainty-calibrated evidential multi-task learning, we propose a novel evidential depth loss function that jointly optimizes the belief strength of the depth prediction in conjunction with evidential segmentation loss. Building on this, we present EvidKimera, an uncertainty-aware semantic surface mapping framework, which uses evidential depth and semantics prediction for improved 3D metric-semantic consistency. We train and evaluate EvidMTL on the NYUDepthV2 and assess its zero-shot performance on ScanNetV2, demonstrating superior uncertainty estimation compared to conventional approaches while maintaining comparable depth estimation and semantic segmentation. In zero-shot mapping tests on ScanNetV2, EvidKimera outperforms Kimera in semantic surface mapping accuracy and consistency, highlighting the benefits of uncertainty-aware mapping and underscoring its potential for real-world robotic applications.
We present Vinci, a vision-language system designed to provide real-time, comprehensive AI assistance on portable devices. At its core, Vinci leverages EgoVideo-VL, a novel model that integrates an egocentric vision foundation model with a large language model (LLM), enabling advanced functionalities such as scene understanding, temporal grounding, video summarization, and future planning. To enhance its utility, Vinci incorporates a memory module for processing long video streams in real time while retaining contextual history, a generation module for producing visual action demonstrations, and a retrieval module that bridges egocentric and third-person perspectives to provide relevant how-to videos for skill acquisition. Unlike existing systems that often depend on specialized hardware, Vinci is hardware-agnostic, supporting deployment across a wide range of devices, including smartphones and wearable cameras. In our experiments, we first demonstrate the superior performance of EgoVideo-VL on multiple public benchmarks, showcasing its vision-language reasoning and contextual understanding capabilities. We then conduct a series of user studies to evaluate the real-world effectiveness of Vinci, highlighting its adaptability and usability in diverse scenarios. We hope Vinci can establish a new framework for portable, real-time egocentric AI systems, empowering users with contextual and actionable insights. Including the frontend, backend, and models, all codes of Vinci are available at https://github.com/OpenGVLab/vinci.
Salient object detection (SOD) in RGB-D images is an essential task in computer vision, enabling applications in scene understanding, robotics, and augmented reality. However, existing methods struggle to capture global dependency across modalities, lack comprehensive saliency priors from both RGB and depth data, and are ineffective in handling low-quality depth maps. To address these challenges, we propose SSNet, a saliency-prior and state space model (SSM)-based network for the RGB-D SOD task. Unlike existing convolution- or transformer-based approaches, SSNet introduces an SSM-based multi-modal multi-scale decoder module to efficiently capture both intra- and inter-modal global dependency with linear complexity. Specifically, we propose a cross-modal selective scan SSM (CM-S6) mechanism, which effectively captures global dependency between different modalities. Furthermore, we introduce a saliency enhancement module (SEM) that integrates three saliency priors with deep features to refine feature representation and improve the localization of salient objects. To further address the issue of low-quality depth maps, we propose an adaptive contrast enhancement technique that dynamically refines depth maps, making them more suitable for the RGB-D SOD task. Extensive quantitative and qualitative experiments on seven benchmark datasets demonstrate that SSNet outperforms state-of-the-art methods.
6D object pose estimation suffers from reduced accuracy when applied to metallic objects. We set out to improve the state-of-the-art by addressing challenges such as reflections and specular highlights in industrial applications. Our novel BOP-compatible dataset, featuring a diverse set of metallic objects (cans, household, and industrial items) under various lighting and background conditions, provides additional geometric and visual cues. We demonstrate that these cues can be effectively leveraged to enhance overall performance. To illustrate the usefulness of the additional features, we improve upon the GDRNPP algorithm by introducing an additional keypoint prediction and material estimator head in order to improve spatial scene understanding. Evaluations on the new dataset show improved accuracy for metallic objects, supporting the hypothesis that additional geometric and visual cues can improve learning.