LARGE-Aircraft
收藏资源简介:
LARGE-Aircraft是第一个专为宽幅图像设计的大规模超高分辨率(VHR)数据集。该数据集包含1,189张大型图像,尺寸从8,000×8,000到20,000×20,000像素不等,总计超过1430亿像素,覆盖全球160多个地点,包含65种细粒度飞机模型。该数据集的核心优势在于其对宽幅图像和细粒度分类的双重强调,与真实卫星观测场景紧密对齐。数据集采用三级层次结构标注,包含10个功能类别的65种特定飞机模型,包括民用飞机、轰炸机、战斗机和预警机等。与传统图像块不同,该数据集中的全场景图像不仅计算成本更高,还引入了复杂的背景杂波和多样的目标分布,这对检测算法的架构和性能提出了更严格的要求。
LARGE-Aircraft is the first large-scale very high-resolution (VHR) dataset specifically designed for wide-format images. This dataset comprises 1,189 large-scale images, with dimensions ranging from 8,000×8,000 to 20,000×20,000 pixels, totaling over 143 billion pixels. It covers more than 160 locations across the globe and includes 65 fine-grained aircraft models. The core advantage of this dataset lies in its dual emphasis on wide-format images and fine-grained classification, which closely aligns with real satellite observation scenarios. The dataset adopts a three-level hierarchical annotation structure, containing 65 specific aircraft models under 10 functional categories, including civil aircraft, bombers, fighters, airborne early warning aircraft, and so on. Unlike traditional image patches, the full-scene images in this dataset not only incur higher computational costs but also introduce complex background clutter and diverse target distributions, which impose more stringent requirements on the architecture and performance of detection algorithms.
LARGE-Aircraft 数据集概述
数据集基本信息
- 数据集名称:LARGE-Aircraft (A Dataset of Large-Size VHR Satellite Imagery for Fine-Grained Aircraft Detection)
- 核心特点:首个专为宽幅影像设计的大规模甚高分辨率卫星图像数据集,强调宽幅影像与细粒度分类。
- 图像数量:1,189 幅大型图像。
- 图像尺寸范围:8,000×8,000 至 20,000×20,000 像素。
- 总像素数:超过 1433 亿像素。
- 覆盖范围:超过 160 个全球地点。
- 标注类别:包含 65 种细粒度飞机型号。
数据集详细描述
- 设计目的:为细粒度飞机检测提供首个大型卫星图像基准数据集。
- 图像内容:包含 1,189 幅机场全场景图像,非传统图像块。
- 数据挑战:
- 计算成本更高。
- 包含复杂的背景杂波。
- 目标分布多样。
- 标注体系:采用三级层次结构进行标注。
- 飞机类别:涵盖 10 个功能类别,包括民用飞机、轰炸机、战斗机、预警机等,来自多个国家。
数据集资源与许可
- 获取状态:文章发表后可用。
- 使用许可:未在提供内容中明确说明。




