AdaptVision: Dynamic Input Scaling in MLLMs for Versatile Scene Understanding

TL;DR

AdaptVision introduces a dynamic input scaling method for multimodal large language models, enabling efficient and accurate scene understanding across diverse image resolutions and content types by adjusting visual tokens based on image complexity.

Contribution

We propose a novel dynamic image partitioning module that adjusts visual tokens according to image size and content, improving versatility and performance in vision-language tasks.

Findings

Effective processing of images up to 1008x1008 resolution.

Improved performance on vision-language tasks across natural and text-rich scenes.

Mitigation of distortion effects from image resizing.

Abstract

Over the past few years, the advancement of Multimodal Large Language Models (MLLMs) has captured the wide interest of researchers, leading to numerous innovations to enhance MLLMs' comprehension. In this paper, we present AdaptVision, a multimodal large language model specifically designed to dynamically process input images at varying resolutions. We hypothesize that the requisite number of visual tokens for the model is contingent upon both the resolution and content of the input image. Generally, natural images with a lower information density can be effectively interpreted by the model using fewer visual tokens at reduced resolutions. In contrast, images containing textual content, such as documents with rich text, necessitate a higher number of visual tokens for accurate text interpretation due to their higher information density. Building on this insight, we devise a dynamic image partitioning module that adjusts the number of visual tokens according to the size and aspect ratio of images. This method mitigates distortion effects that arise from resizing images to a uniform resolution and dynamically optimizing the visual tokens input to the LLMs. Our model is capable of processing images with resolutions up to $1008\times 1008$. Extensive experiments across various datasets demonstrate that our method achieves impressive performance in handling vision-language tasks in both natural and text-related scenes. The source code and dataset are now publicly available at \url{https://github.com/harrytea/AdaptVision}.