Language-Image Models with 3D Understanding
read the original abstract
Multi-modal large language models (MLLMs) have shown incredible capabilities in a variety of 2D vision and language tasks. We extend MLLMs' perceptual capabilities to ground and reason about images in 3-dimensional space. To that end, we first develop a large-scale pre-training dataset for 2D and 3D called LV3D by combining multiple existing 2D and 3D recognition datasets under a common task formulation: as multi-turn question-answering. Next, we introduce a new MLLM named Cube-LLM and pre-train it on LV3D. We show that pure data scaling makes a strong 3D perception capability without 3D specific architectural design or training objective. Cube-LLM exhibits intriguing properties similar to LLMs: (1) Cube-LLM can apply chain-of-thought prompting to improve 3D understanding from 2D context information. (2) Cube-LLM can follow complex and diverse instructions and adapt to versatile input and output formats. (3) Cube-LLM can be visually prompted such as 2D box or a set of candidate 3D boxes from specialists. Our experiments on outdoor benchmarks demonstrate that Cube-LLM significantly outperforms existing baselines by 21.3 points of AP-BEV on the Talk2Car dataset for 3D grounded reasoning and 17.7 points on the DriveLM dataset for complex reasoning about driving scenarios, respectively. Cube-LLM also shows competitive results in general MLLM benchmarks such as refCOCO for 2D grounding with (87.0) average score, as well as visual question answering benchmarks such as VQAv2, GQA, SQA, POPE, etc. for complex reasoning. Our project is available at https://janghyuncho.github.io/Cube-LLM.
This paper has not been read by Pith yet.
Forward citations
Cited by 6 Pith papers
-
Abstract 3D Perception for Spatial Intelligence in Vision-Language Models
SandboxVLM enhances VLMs' spatial intelligence by encoding 3D geometry with abstract bounding boxes in a four-stage zero-shot pipeline, yielding an 8.3% improvement on SAT Real benchmark.
-
GeoWorld-VLM: Geometry from World Models for Vision-Language Models
GeoWorld-VLM aligns VLM image features with intermediate representations from camera-conditioned world models via fine-tuning only the encoder and projector, yielding ~4% gains on What'sUp and VSR spatial benchmarks a...
-
Boxer: Robust Lifting of Open-World 2D Bounding Boxes to 3D
BoxerNet lifts 2D bounding boxes to metric 3D boxes via transformer regression with aleatoric uncertainty and median depth encoding, then fuses multi-view results to outperform CuTR by large margins on open-world benchmarks.
-
Universal Pose Pretraining for Generalizable Vision-Language-Action Policies
Pose-VLA uses a decoupled two-stage pre-training with discrete pose tokens to extract universal 3D spatial priors from 3D datasets and robotic trajectories, achieving 79.5% success on RoboTwin 2.0 and 96.0% on LIBERO.
-
EMMA: End-to-End Multimodal Model for Autonomous Driving
EMMA is an end-to-end multimodal LLM that converts camera data into trajectories, objects, and road graphs via text prompts and reports state-of-the-art motion planning on nuScenes plus competitive detection results on Waymo.
-
GeoWorld-VLM: Geometry from World Models for Vision-Language Models
GeoWorld-VLM distills geometric structure from camera-conditioned world models into VLMs by aligning visual features, improving spatial reasoning by about 4% on What'sUp and VSR benchmarks across two architectures whi...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.