UHR-BAT: Budget-Aware Token Compression Vision-Language model for Ultra-High-Resolution Remote Sensing

TL;DR

UHR-BAT is a novel query-guided token compression framework for ultra-high-resolution remote sensing images, balancing detail preservation and computational efficiency using multi-scale importance estimation and region-wise strategies.

Contribution

It introduces a region-faithful, query-guided token compression method that improves efficiency without sacrificing critical image details in UHR remote sensing tasks.

Findings

Achieves state-of-the-art performance on multiple benchmarks.

Effectively reduces computational cost while maintaining detail accuracy.

Utilizes multi-scale importance estimation for precise token selection.

Abstract

Ultra-high-resolution (UHR) remote sensing imagery couples kilometer-scale context with query-critical evidence that may occupy only a few pixels. Such vast spatial scale leads to a quadratic explosion of visual tokens and hinders the extraction of information from small objects. Previous works utilize direct downsampling, dense tiling, or global top-k pruning, which either compromise query-critical image details or incur unpredictable compute. In this paper, we propose UHR-BAT, a query-guided and region-faithful token compression framework to efficiently select visual tokens under a strict context budget. Specifically, we leverage text-guided, multi-scale importance estimation for visual tokens, effectively tackling the challenge of achieving precise yet low-cost feature extraction. Furthermore, by introducing region-wise preserve and merge strategies, we mitigate visual token redundancy, further driving down the computational budget. Experimental results show that UHR-BAT achieves state-of-the-art performance across various benchmarks. Code will be available at https://github.com/Yunkaidang/UHR.