GPT4Scene: Understand 3D Scenes from Videos with Vision-Language Models
Reviewed by Pithpith:3VDRO65Gopen to challenge →
read the original abstract
In recent years, 2D Vision-Language Models (VLMs) have made significant strides in image-text understanding tasks. However, their performance in 3D spatial comprehension, which is critical for embodied intelligence, remains limited. Recent advances have leveraged 3D point clouds and multi-view images as inputs, yielding promising results. However, we propose exploring a purely vision-based solution inspired by human perception, which merely relies on visual cues for 3D spatial understanding. This paper empirically investigates the limitations of VLMs in 3D spatial knowledge, revealing that their primary shortcoming lies in the lack of global-local correspondence between the scene and individual frames. To address this, we introduce GPT4Scene, a novel visual prompting paradigm in VLM training and inference that helps build the global-local relationship, significantly improving the 3D spatial understanding of indoor scenes. Specifically, GPT4Scene constructs a Bird's Eye View (BEV) image from the video and marks consistent object IDs across both frames and the BEV image. The model then inputs the concatenated BEV image and video frames with markers. In zero-shot evaluations, GPT4Scene improves performance over closed-source VLMs like GPT-4o. Additionally, we prepare a processed video dataset consisting of 165K text annotation to fine-tune open-source VLMs, achieving state-of-the-art performance on all 3D understanding tasks. Surprisingly, after training with the GPT4Scene paradigm, VLMs consistently improve during inference, even without object marker prompting and BEV image as explicit correspondence. It demonstrates that the proposed paradigm helps VLMs develop an intrinsic ability to understand 3D scenes, which paves the way for a seamless approach to extending pre-trained VLMs for 3D scene understanding.
This paper has not been read by Pith yet.
Forward citations
Cited by 28 Pith papers
-
ViSRA: A Video-based Spatial Reasoning Agent for Multi-modal Large Language Models
ViSRA boosts MLLM 3D spatial reasoning performance by up to 28.9% on unseen tasks via a plug-and-play video-based agent that extracts explicit spatial cues from expert models without any post-training.
-
SpatialBench: Benchmarking Multimodal Large Language Models for Spatial Cognition
SpatialBench creates a five-level framework and 15-task benchmark to measure hierarchical spatial reasoning in MLLMs, finding strong basic perception but weak symbolic reasoning, causal inference, and planning.
-
CAIRN: Cross-Room 3D Scene Understanding with Topology-Aware Large Multimodal Models
A topology-aware 3D-LLM with hierarchical masked attention and geometric bias outperforms prior 3D-LLMs on a new multi-room scene understanding benchmark built from HM3D.
-
Agentic Collaborative Cognition for Zero-Shot 3D Understanding
A collaborative Planning-Perception agent framework using MLLMs constructs a holistic cognitive map through iterative viewpoint supplementation and achieves reported SOTA gains on six 3D benchmarks.
-
Agentic Collaborative Cognition for Zero-Shot 3D Understanding
A closed-loop multi-agent framework with Planning and Perception agents iteratively supplements viewpoints and integrates object observations into a holistic cognitive map, achieving SOTA on six 3D benchmarks.
-
Occ-VLM: Occupancy Grounded Vision Language Model for Indoor Scene Understanding
Occ-VLM reconstructs 3D occupancy from 2D images via a single encoder to ground vision-language reasoning, claiming SOTA occupancy prediction and parity with 3D-input VLMs on VQA and captioning.
-
Stream3D-VLM: Online 3D Spatial Understanding with Incremental Geometry Priors
Stream3D-VLM adds autoregressive streaming control, VSFI geometry integration, GAVC compression, and a 1M-pair benchmark to enable real-time 3D VLM performance that beats prior models on 29 online and offline tasks.
-
Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning
GASP injects geometric priors into VLMs via a deep-supervised correspondence head trained on video point correspondences and depth consistency, raising internal matching accuracy and delivering gains on spatial benchm...
-
SSR3D-LLM: Structured Spatial Reasoning via Latent Steps for Fine-Grained Grounding in Unified 3D-LLMs
SSR3D-LLM improves fine-grained 3D grounding in unified 3D-LLMs by generating and scoring sequences of latent spatial reasoning steps from the query using fixed Mask3D proposals.
-
Bridging the 2D-3D Gap: A Hierarchical Semantic-Geometric Map for Vision Language Navigation
HSGM structures 3D geometry and semantics into a multi-level map that lets VLMs perform high-level planning in zero-shot VLN, achieving SOTA on R2R-CE and RxR-CE.
-
EgoProx: Evaluating MLLMs on Egocentric 3D Proximity Reasoning Across a Cognitive Hierarchy
EgoProx benchmark shows MLLMs have some spatial knowledge but struggle to leverage it for egocentric 3D proximity reasoning VQA.
-
Unlocking Dense Metric Depth Estimation in VLMs
DepthVLM converts a standard VLM into a dense metric depth predictor by attaching a lightweight head and training under unified vision-text supervision, outperforming prior VLMs and some pure vision models on a new in...
-
Unlocking Dense Metric Depth Estimation in VLMs
DepthVLM attaches a depth head to VLMs for native dense metric depth prediction alongside language outputs using a two-stage unified training schedule and a new indoor-outdoor benchmark.
-
Multi-Scale Gaussian-Language Map for Zero-shot Embodied Navigation and Reasoning
GLMap combines explicit 3D Gaussians with multi-scale language semantics in a dual-modality structure and uses an analytical Gaussian Estimator for incremental map building, improving zero-shot performance on navigati...
-
Geometry-Guided 3D Visual Token Pruning for Video-Language Models
Geo3DPruner uses geometry-aware global attention and two-stage voxel pruning to remove 90% of visual tokens from spatial videos while keeping over 90% of original performance on 3D scene benchmarks.
-
Let Geometry GUIDE: Layer-wise Unrolling of Geometric Priors in Multimodal LLMs
GUIDE unrolls multi-granularity geometric priors layer-wise into early MLLM layers with gating to improve spatial reasoning and perception.
-
EgoMind: Activating Spatial Cognition through Linguistic Reasoning in MLLMs
EgoMind activates spatial cognition in MLLMs via linguistic Role-Play Caption and Progressive Spatial Analysis, reaching competitive results on VSI-Bench, SPAR-Bench, SITE-Bench and SPBench with only 5K SFT and 20K RL...
-
Feeling the Space: Egomotion-Aware Video Representation for Efficient and Accurate 3D Scene Understanding
Motion-MLLM integrates IMU egomotion data into MLLMs using cascaded filtering and asymmetric fusion to ground visual content in physical trajectories for scale-aware 3D understanding, achieving competitive accuracy at...
-
Boosting MLLM Spatial Reasoning with Geometrically Referenced 3D Scene Representations
GR3D turns 3D scene geometry into ID-indexed text references, enabling zero-shot MLLM spatial reasoning gains of 9% on VSI-Bench and 12% on MindCube.
-
Reinforcing Spatial Reasoning in Vision-Language Models with Interwoven Thinking and Visual Drawing
VILASR integrates visual drawing operations with reasoning in LVLMs via cold-start synthetic training, reflective rejection sampling, and reinforcement learning, yielding an 18.4% average gain on spatial reasoning benchmarks.
-
Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence
Spatial-MLLM adds a 3D spatial encoder initialized from a visual geometry model and space-aware frame sampling to MLLMs to improve spatial understanding and reasoning from purely 2D visual inputs.
-
Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence
Spatial-MLLM boosts MLLM spatial intelligence from 2D inputs via dual encoders initialized from geometry models plus space-aware sampling, claiming state-of-the-art results.
-
ReScene: Structured Indoor Scene Reconstruction from Multi-View Captures
ReScene introduces HierView for view prioritization and Relation-Aware Assembly for scene graph fusion, reporting 17% lower Chamfer Distance and 26% lower LPIPS than prior baselines on ScanNet while running faster.
-
SpatialSV: Internalizing Interpretable 3D Spatial Awareness in MLLMs via Task-Oriented Visual Supervision
SpatialSV trains MLLMs to lift 2D visual features into explicit 3D representations via task-oriented supervision for interpretable spatial awareness.
-
AgentGrounder: Zero-Shot 3D Visual Pointcloud Grounding using Multimodal Language Models
AgentGrounder performs zero-shot 3D visual grounding on colored point clouds via an offline object lookup table and an online agent that selectively retrieves, scores geometrically, and renders images on demand, repor...
-
Efficient3D: A Unified Framework for Adaptive and Debiased Token Reduction in 3D MLLMs
Efficient3D prunes visual tokens in 3D MLLMs via DVTIE and ATR modules, reporting better performance than unpruned baselines on Scan2Cap and other benchmarks.
-
3D-IDE: 3D Implicit Depth Emergent
3D awareness emerges implicitly in MLLMs via self-supervised geometric constraints that create an information bottleneck, removing depth and pose dependencies at inference and cutting latency by 55%.
-
SpaceEra++: A Unified Framework Towards 3D Spatial Reasoning in Video
SpaceEra++ adds ScenePick frame sampling and SpaceAlign pairwise constraints to the prior SpaceEra system, claiming consistent benchmark gains for 3D video spatial reasoning.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.