LLaVA-VSD: Large Language-and-Vision Assistant for Visual Spatial Description

TL;DR

LLaVA-VSD is a large multimodal model that generates detailed and accurate visual spatial descriptions by combining vision and language understanding, fine-tuned on a specialized dataset for spatial relationship tasks.

Contribution

The paper introduces LLaVA-VSD, a novel large language-and-vision model specifically designed for visual spatial description, integrating high-resolution image understanding and open-ended language capabilities.

Findings

Effective classification and description of visual spatial relationships.

Supports high-resolution images and open-ended spatial inquiries.

Enhanced sentence diversity and accuracy through language model refinement.

Abstract

Visual Spatial Description (VSD) aims to generate texts that describe the spatial relationships between objects within images. Traditional visual spatial relationship classification (VSRC) methods typically output the spatial relationship between two objects in an image, often neglecting world knowledge and lacking general language capabilities. In this paper, we propose a Large Language-and-Vision Assistant for Visual Spatial Description, named LLaVA-VSD, which is designed for the classification, description, and open-ended description of visual spatial relationships. Specifically, the model first constructs a VSD instruction-following dataset using given figure-caption pairs for the three tasks. It then employs LoRA to fine-tune a Large Language and Vision Assistant for VSD, which has 13 billion parameters and supports high-resolution images. Finally, a large language model (Qwen-2) is used to refine the generated sentences, enhancing their diversity and accuracy. LLaVA-VSD demonstrates excellent multimodal conversational capabilities and can follow open-ended instructions to assist with inquiries about object relationships in images.