DenseWorld-1M: Towards Detailed Dense Grounded Caption in the Real World

TL;DR

DenseWorld-1M is a large-scale, detailed dataset with dense grounded captions for real-world scenes, designed to enhance multimodal models' understanding of complex visual relations and object descriptions.

Contribution

We introduce DenseWorld-1M, the first large-scale dataset with detailed dense grounded captions, and develop a three-stage labeling pipeline with specialized VLM models to improve caption quality.

Findings

DenseWorld-1M improves scene understanding in vision-language tasks.

Our models enhance caption detail and spatial relation accuracy.

Experiments show significant gains in visual grounding and caption generation.

Abstract

Multimodal Large Language Models (MLLMs) demonstrate a complex understanding of scenes, benefiting from large-scale and high-quality datasets. Most existing caption datasets lack the ground locations and relations for visual entities. Several grounded caption datasets face the problems of missing detailed descriptions, relations, and massive object descriptions on high-resolution images. To fill this gap for the community, we present DenseWorld-1M, the first massive, detailed, dense grounded caption dataset in the real world. We design a three-stage labeling pipeline, containing open-world perception, detailed object caption generation, and dense caption merging. The first stage obtains entity-level masks and labels. The second stage generates the object-level, detailed captions with the guidance of masks and labels from the first stage. The final stage merges object captions and masks into spatial and relational dense captions. To accelerate the labeling process and improve caption quality, we present two VLM models: the Detailed Region Caption model and the Spatial Caption Merging model. Extensive experiments on various settings, including vision-language understanding, visual grounding, and region caption generation, demonstrate the effectiveness of our DenseWorld-1M dataset and labeling models.