CVMar 20, 2025

DynamicVis: An Efficient and General Visual Foundation Model for Remote Sensing Image Understanding

arXiv:2503.16426v119 citationsh-index: 41
Originality Incremental advance
AI Analysis

This addresses the challenge of efficiently modeling cross-task knowledge from high-resolution remote sensing imagery where targets occupy minimal spatial proportions (~1%), though it appears incremental as an adaptation of existing foundation model concepts to a specific domain.

The paper tackles the problem of limited generalization in remote sensing image understanding by proposing DynamicVis, a foundation model that achieves 97 ms latency (6% of ViT's) and 833 MB GPU memory (3% of ViT's) while processing 2048x2048 pixel images.

The advancement of remote sensing technology has improved the spatial resolution of satellite imagery, facilitating more detailed visual representations for diverse interpretations. However, existing methods exhibit limited generalization capabilities across varied applications. While some contemporary foundation models demonstrate potential, they are hindered by insufficient cross-task adaptability and primarily process low-resolution imagery of restricted sizes, thus failing to fully exploit high-resolution data or leverage comprehensive large-scene semantics. Crucially, remote sensing imagery differs fundamentally from natural images, as key foreground targets (eg., maritime objects, artificial structures) often occupy minimal spatial proportions (~1%) and exhibit sparse distributions. Efficiently modeling cross-task generalizable knowledge from lengthy 2D tokens (~100,000) poses a significant challenge yet remains critical for remote sensing image understanding. Motivated by the selective attention mechanisms inherent to the human visual system, we propose DynamicVis, a dynamic visual perception foundation model for remote sensing imagery. The framework integrates a novel dynamic region perception backbone based on the selective state space model, which strategically balances localized detail extraction with global contextual integration, enabling computationally efficient encoding of large-scale data while maintaining architectural scalability. To enhance cross-task knowledge transferring, we introduce a multi-instance learning paradigm utilizing meta-embedding representations, trained on million-scale region-level annotations. Evaluations across nine downstream tasks demonstrate the model's versatility. DynamicVis achieves multi-level feature modeling with exceptional efficiency, processing (2048x2048) pixels with 97 ms latency (6% of ViT's) and 833 MB GPU memory (3% of ViT's).

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes