Skip to content
Preprint

Understanding Dynamic Scenes at Gigapixel Scale: Wide-Area Spatio-Temporal Perception from UAVs

Sep 2026 · 0 citations · 46 references
Computer Science

TL;DR

This work proposes a latency-aware metric for multi-object tracking called streaming-HOTA (s-HOTA), and introduces an ultra-High-resolution Airborne Remote-sensing Dataset (HARD) annotated at three levels for object detection, multi-object tracking, and scene-level visual question answering.

Abstract

UAV-borne imaging has advanced from megapixel to gigapixel sensors, shifting aerial perception from recognizing individual targets to understanding entire dynamic scenes. We characterize this demand as Wide-area Spatio-temporal Scene Understanding (WSTU), which requires wide-area coverage, per-target resolution, and temporal continuity at once, a combination existing datasets lack. To fill this gap, we introduce an ultra-High-resolution (12768x9564) Airborne Remote-sensing Dataset (HARD) annotated at three levels for object detection, multi-object tracking, and scene-level visual question answering. Ultra-high-resolution imagery raises per-frame processing time to seconds. At that scale latency can no longer be ignored in evaluation. Thus, we propose a latency-aware metric for multi-object tracking called streaming-HOTA (s-HOTA). Extensive baseline experiments show how ultra-high-resolution processing reshapes each task. For detection, the end-to-end pipeline affects accuracy and speed as much as the detector itself does. For tracking, high latency charges the association axis far more unevenly than the detection axis, and association is where pipelines diverge. As a result, the pipeline that performs best offline can lose its lead under s-HOTA. For VQA, vision-language models remain weak at cross-frame identity binding and cannot transfer their single-frame gains to it. Together these findings show that the baselines we evaluate fall short of WSTU. HARD provides the data and the systematic baselines to advance it.

View source

Similar papers

2026

FSM: Frequency-Enhanced Spatiotemporal Model for Streaming Remote Sensing Object Detection

Remote sensing object detection is a fundamental task in ground scene observation and analysis. Despite the currently discrete-frame detectors achieves remarkable performance, they still suffer from three critical limitations: 1) mainstream architectures regress spatial locations independently, making it difficult to e...

Shi-Long Jing, Heng-Yi Lv, Yu-Chen Zhao et al. · 0 citations
Sep 2026

AevaScenes: An FMCW LiDAR Dataset and Benchmark for Long-Range Perception

FMCW LiDAR measures per-point radial Doppler velocity alongside range, providing a motion cue unavailable in conventional time-of-flight sensors. Exploiting this signal at long range remains understudied. We present an FMCW LiDAR dataset of 575 sequences (57.5K frames) with over 8 million annotated 3D boxes across 16 d...

G. Narasimhan, Heethesh Vhavle, Kumar Bhargav Viswanatha et al. · 1 citation
Open access Sep 2026

P2LITE-YOLO: DETECTION-SCALE REALLOCATION FOR TINY OBJECT DETECTION IN UAV IMAGERY

In UAV imagery, object scales are heavily biased towards the tiny region, yet general-purpose object detectors typically employ fixed multi-scale prediction configurations. There is a significant mismatch between the spatial detail loss caused by continuous downsampling and the prediction scale setup. To address this i...

Kai-Qi Mi · 0 citations
Open access 2026

SL-YOLO: Scale- and Lighting-Aware Object Detection for Challenging UAV Imagery

Complex illumination, perspective-induced scale variation, and repeated downsampling degrade object representations in unmanned aerial vehicle (UAV) imagery and make small-object localization highly sensitive to pixel-level shifts. To address these issues, we propose Scale- and Lighting-Aware YOLO (SL-YOLO), a lightwei...

Heng-Sheng Cai, Yuan-Lin Wang, Li Zhao et al. · 0 citations
#graph neural networks Open access Oct 2026

Trajectory-SAM: Object Trajectory-Guided Segmentation for Remote Sensing Videos

The rapid proliferation of high-resolution remote sensing video (RSV) has made object segmentation an increasingly critical yet challenging task. Existing methods face two fundamental limitations in this domain: (1) the intrinsic complexity of RSVs arising from diminutive object sizes, dense distributions, and drastic...

Shi-Yao Guo, Cengiz Ozel, Yang Qu · 0 citations
Open access Sep 2026

VesselBench-800K: A Large-scale Perception Benchmark for Multimodal Vessel Detection, Counting, and Density Estimation

VesselBench-800K, the largest-to-date benchmark dataset on a global scale for vessel perception in multimodal RS images, comprises 800,000 images, each at a resolution of 512x512 pixels, specifically curated for vessel perception tasks such as detection, counting, and density estimation.

D. Hong, Chen-Yu Li, J. Chanussot · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.