RoboSpatial: Teaching Spatial Understanding to 2D and 3D Vision-Language Models for Robotics

• URL: http://arxiv.org/abs/2411.16537v4
• Date: Sat, 05 Apr 2025 06:46:03 GMT
• Title: RoboSpatial: Teaching Spatial Understanding to 2D and 3D Vision-Language Models for Robotics
• Authors: Chan Hee Song, Valts Blukis, Jonathan Tremblay, Stephen Tyree, Yu Su, Stan Birchfield,
• Abstract summary: We introduce RoboSpatial, a large-scale dataset for spatial understanding in robotics.<n>It consists of real indoor and tabletop scenes, captured as 3D scans and egocentric images, and annotated with rich spatial information relevant to robotics.<n>Our experiments show that models trained with RoboSpatial outperform baselines on downstream tasks such as spatial affordance prediction, spatial relationship prediction, and robot manipulation.
• Score: 26.42651735582044
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Spatial understanding is a crucial capability that enables robots to perceive their surroundings, reason about their environment, and interact with it meaningfully. In modern robotics, these capabilities are increasingly provided by vision-language models. However, these models face significant challenges in spatial reasoning tasks, as their training data are based on general-purpose image datasets that often lack sophisticated spatial understanding. For example, datasets frequently do not capture reference frame comprehension, yet effective spatial reasoning requires understanding whether to reason from ego-, world-, or object-centric perspectives. To address this issue, we introduce RoboSpatial, a large-scale dataset for spatial understanding in robotics. It consists of real indoor and tabletop scenes, captured as 3D scans and egocentric images, and annotated with rich spatial information relevant to robotics. The dataset includes 1M images, 5k 3D scans, and 3M annotated spatial relationships, and the pairing of 2D egocentric images with 3D scans makes it both 2D- and 3D- ready. Our experiments show that models trained with RoboSpatial outperform baselines on downstream tasks such as spatial affordance prediction, spatial relationship prediction, and robot manipulation.

Related papers

• Lift3D Foundation Policy: Lifting 2D Large-Scale Pretrained Models for Robust 3D Robotic Manipulation [30.744137117668643]
  Lift3D is a framework that enhances 2D foundation models with implicit and explicit 3D robotic representations to construct a robust 3D manipulation policy. In experiments, Lift3D consistently outperforms previous state-of-the-art methods across several simulation benchmarks and real-world scenarios.
  arXiv  Detail & Related papers  (2024-11-27T18:59:52Z)
• Grounding 3D Scene Affordance From Egocentric Interactions [52.5827242925951]
  Grounding 3D scene affordance aims to locate interactive regions in 3D environments. We introduce a novel task: grounding 3D scene affordance from egocentric interactions.
  arXiv  Detail & Related papers  (2024-09-29T10:46:19Z)
• SUGAR: Pre-training 3D Visual Representations for Robotics [85.55534363501131]
  We introduce a novel 3D pre-training framework for robotics named SUGAR. SUGAR captures semantic, geometric and affordance properties of objects through 3D point clouds. We show that SUGAR's 3D representation outperforms state-of-the-art 2D and 3D representations.
  arXiv  Detail & Related papers  (2024-04-01T21:23:03Z)
• Teaching Unknown Objects by Leveraging Human Gaze and Augmented Reality in Human-Robot Interaction [3.1473798197405953]
  This dissertation aims to teach a robot unknown objects in the context of Human-Robot Interaction (HRI) The combination of eye tracking and Augmented Reality created a powerful synergy that empowered the human teacher to communicate with the robot. The robot's object detection capabilities exhibited comparable performance to state-of-the-art object detectors trained on extensive datasets.
  arXiv  Detail & Related papers  (2023-12-12T11:34:43Z)
• Uncertainty-aware State Space Transformer for Egocentric 3D Hand Trajectory Forecasting [79.34357055254239]
  Hand trajectory forecasting is crucial for enabling a prompt understanding of human intentions when interacting with AR/VR systems. Existing methods handle this problem in a 2D image space which is inadequate for 3D real-world applications. We set up an egocentric 3D hand trajectory forecasting task that aims to predict hand trajectories in a 3D space from early observed RGB videos in a first-person view.
  arXiv  Detail & Related papers  (2023-07-17T04:55:02Z)
• A Universal Semantic-Geometric Representation for Robotic Manipulation [42.18087956844491]
  We present $textbfSemantic-Geometric Representation (textbfSGR)$, a universal perception module for robotics. SGR leverages the rich semantic information of large-scale pre-trained 2D models and inherits the merits of 3D spatial reasoning. Our experiments demonstrate that SGR empowers the agent to successfully complete a diverse range of simulated and real-world robotic manipulation tasks.
  arXiv  Detail & Related papers  (2023-06-18T04:34:17Z)
• ScanERU: Interactive 3D Visual Grounding based on Embodied Reference Understanding [67.21613160846299]
  Embodied Reference Understanding (ERU) is first designed for this concern. New dataset called ScanERU is constructed to evaluate the effectiveness of this idea.
  arXiv  Detail & Related papers  (2023-03-23T11:36:14Z)
• CLIP$^2$: Contrastive Language-Image-Point Pretraining from Real-World Point Cloud Data [80.42480679542697]
  We propose Contrastive Language-Image-Point Cloud Pretraining (CLIP$2$) to learn the transferable 3D point cloud representation in realistic scenarios. Specifically, we exploit naturally-existed correspondences in 2D and 3D scenarios, and build well-aligned and instance-based text-image-point proxies from those complex scenarios.
  arXiv  Detail & Related papers  (2023-03-22T09:32:45Z)
• Neural Scene Representation for Locomotion on Structured Terrain [56.48607865960868]
  We propose a learning-based method to reconstruct the local terrain for a mobile robot traversing urban environments. Using a stream of depth measurements from the onboard cameras and the robot's trajectory, the estimates the topography in the robot's vicinity. We propose a 3D reconstruction model that faithfully reconstructs the scene, despite the noisy measurements and large amounts of missing data coming from the blind spots of the camera arrangement.
  arXiv  Detail & Related papers  (2022-06-16T10:45:17Z)
• Extracting Zero-shot Common Sense from Large Language Models for Robot 3D Scene Understanding [25.270772036342688]
  We introduce a novel method for leveraging common sense embedded within large language models for labelling rooms. The proposed algorithm operates on 3D scene graphs produced by modern spatial perception systems.
  arXiv  Detail & Related papers  (2022-06-09T16:05:35Z)
• 3D Neural Scene Representations for Visuomotor Control [78.79583457239836]
  We learn models for dynamic 3D scenes purely from 2D visual observations. A dynamics model, constructed over the learned representation space, enables visuomotor control for challenging manipulation tasks.
  arXiv  Detail & Related papers  (2021-07-08T17:49:37Z)
• LanguageRefer: Spatial-Language Model for 3D Visual Grounding [72.7618059299306]
  We develop a spatial-language model for a 3D visual grounding problem. We show that our model performs competitively on visio-linguistic datasets proposed by ReferIt3D.
  arXiv  Detail & Related papers  (2021-07-07T18:55:03Z)
• Task-relevant Representation Learning for Networked Robotic Perception [74.0215744125845]
  This paper presents an algorithm to learn task-relevant representations of sensory data that are co-designed with a pre-trained robotic perception model's ultimate objective. Our algorithm aggressively compresses robotic sensory data by up to 11x more than competing methods.
  arXiv  Detail & Related papers  (2020-11-06T07:39:08Z)
• Learning Object Placements For Relational Instructions by Hallucinating Scene Representations [26.897316325189205]
  We present a convolutional neural network for estimating pixelwise object placement probabilities for a set of spatial relations from a single input image. Our method does not require ground truth data for the pixelwise relational probabilities or 3D models of the objects. Results obtained using real-world data and human-robot experiments demonstrate the effectiveness of our method.
  arXiv  Detail & Related papers  (2020-01-23T12:58:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.