The All-Seeing Project V2: Towards General Relation Comprehension of the Open World

TL;DR

The paper introduces ASMv2, a model and dataset for comprehensive relation understanding in images, improving object and relation recognition and reducing hallucinations in multi-modal models, with new benchmarks and datasets.

Contribution

We propose ASMv2, a unified model for relation comprehension, and introduce the first high-quality ReC dataset and a new evaluation benchmark for relation understanding in MLLMs.

Findings

ASMv2 achieves 52.04% accuracy on CRPE benchmark.

Our dataset aligns with instruction tuning data for training.

ASMv2 outperforms previous models like LLaVA-1.5 significantly.

Abstract

We present the All-Seeing Project V2: a new model and dataset designed for understanding object relations in images. Specifically, we propose the All-Seeing Model V2 (ASMv2) that integrates the formulation of text generation, object localization, and relation comprehension into a relation conversation (ReC) task. Leveraging this unified task, our model excels not only in perceiving and recognizing all objects within the image but also in grasping the intricate relation graph between them, diminishing the relation hallucination often encountered by Multi-modal Large Language Models (MLLMs). To facilitate training and evaluation of MLLMs in relation understanding, we created the first high-quality ReC dataset ({AS-V2) which is aligned with the format of standard instruction tuning data. In addition, we design a new benchmark, termed Circular-based Relation Probing Evaluation (CRPE) for comprehensively evaluating the relation comprehension capabilities of MLLMs. Notably, our ASMv2 achieves an overall accuracy of 52.04 on this relation-aware benchmark, surpassing the 43.14 of LLaVA-1.5 by a large margin. We hope that our work can inspire more future research and contribute to the evolution towards artificial general intelligence. Our project is released at https://github.com/OpenGVLab/all-seeing.