C2A: Human Detection in Disaster Scenarios

该数据集名为C2A，是为了推进灾难场景中的人体检测技术而精心策划的集合。它融合了AIDER数据集中的灾难场景背景图与LSP/MPII-MPHB数据集中的多样化人体姿态。图片尺寸从大约123x152像素到高分辨率的5184x3456像素不等，其中中等图像宽度为428像素。该数据集包含了各种人体大小和姿态的多样性。规模上，它包含了10,215张图像，以及超过36万个用于人体检测的对象。该数据集的任务是针对灾难场景中的人体检测。

This dataset, named C2A, is a meticulously curated collection developed to advance human detection technologies in disaster scenarios. It integrates disaster scenario background images from the AIDER dataset and diverse human poses from the LSP/MPII-MPHB datasets. The image sizes range from approximately 123×152 pixels to the high-resolution 5184×3456 pixels, with the median image width being 428 pixels. This dataset features rich diversity in terms of human body sizes and poses. In terms of scale, it contains 10,215 images and over 360,000 annotated objects for human detection tasks. The core task of this dataset is human detection in disaster scenarios.

搜集汇总

背景与挑战

背景概述

C2A数据集是一个用于灾害场景下人体检测的合成数据集，包含10,215张图像和超过360,000个标注人体实例，涵盖多种人体姿态和灾害场景。数据集提供YOLO和COCO两种标注格式，并包含详细的姿态信息，旨在提升复杂灾害环境中的人体检测性能。