LLaVA-3D: A Simple yet Effective Pathway to Empowering LMMs with 3D-awareness

TL;DR

LLaVA-3D introduces a unified framework that enhances large multimodal models with 3D scene understanding by integrating 3D spatial context, enabling accurate 3D perception without sacrificing 2D capabilities and improving training efficiency.

Contribution

The paper presents a simple, effective method to adapt LLaVA for 3D understanding using 3D position embeddings and joint training, achieving state-of-the-art results and faster convergence.

Findings

Supports direct decoding of 3D spatial outputs like bounding boxes

Converges 3.5x faster than previous 3D LMMs

Maintains comparable 2D understanding and conversation abilities

Abstract

Recent advancements in Large Multimodal Models (LMMs) have greatly enhanced their proficiency in 2D visual understanding tasks, enabling them to effectively process and understand images and videos. However, the development of LMMs with 3D scene understanding capabilities has been hindered by the lack of large-scale 3D vision-language datasets and powerful 3D encoders. In this paper, we introduce a simple yet effective framework called LLaVA-3D. Leveraging the strong 2D visual understanding priors from LLaVA, our LLaVA-3D efficiently adapts LLaVA for 3D scene understanding without compromising 2D understanding capabilities. To achieve this, we utilize the 3D position embeddings to enhance the 2D CLIP Patches with 3D spatial context information and construct 3D patches. By integrating the 3D position embeddings into 2D LMMs and employing joint 2D and 3D vision-language instruction tuning, we establish a unified architecture for both 2D visual understanding and 3D scene understanding. In contrast to previous 3D LMMs, LLaVA-3D supports decoding accurate 3D spatial perception outputs, e.g., 3D bounding boxes, directly from these 3D patches, without relying on the time-consuming off-the-shelf 3D segmentors. Experimental results show that LLaVA-3D converges 3.5x faster than existing 3D LMMs when trained on 3D vision-language datasets. Moreover, LLaVA-3D not only achieves state-of-the-art performance across various 3D tasks but also maintains comparable 2D visual understanding and vision-language conversation capabilities with LLaVA.