logo
CanyamAI summaries for academic research
language
MobileMobile
  • Phone FeatureWeChat Mini Program
EnglishEnglish
noticeNotice
BlogBlog
delete
arrow
Return

DynamicVis: Dynamic visual perception for efficient remote sensing foundation models

delete2026-10-13
delete0
PRE
AI
K
Keyan Chen
C
Chenyang Liu
B
Bowen Chen
W
Wenyuan Li
Z
Zhengxia Zou
LU Shijian cover
LU Shijian (Shijian Lu)
Z
Zhenwei Shi *
DOI：10.1016/j.isprsjprs.2026.09.042delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
• DynamicVis enables efficient sparse spatial perception in remote sensing imagery. • ATRIM selectively routes salient tokens while bypassing redundant backgrounds. • MIL pre-training enhances foreground-aware regional semantic discrimination. • DynamicVis transfers across nine tasks with low latency and GPU memory.
Keywords:
Remote sensing foundation model
Dynamic spatial perception
State space models
Multi-instance learning
Spatial sparsity

Journal

ISPRS Journal of Photogrammetry and Remote Sensing cover
ISPRS Journal of Photogrammetry and Remote Sensing
IF:
12.2
Papers:
4.4K
Citations:
3.2W

Organization

N
nanyang technological university
Scholars:
1.1K
Papers: 566
Citations: 0
B
beihang university
Scholars:
2.8K
Papers: 872
Citations: 0
T
The University of Hong Kong
Scholars:
778
Papers: 357
Citations: 0
researcher View more organizations
Cited Papers

Cited Papers

No cited papers available