北川现场可见光红外数据集

该数据集包含针对北川地震遗址现场场景进行采集三个场景假人和真人不同形态的可见光和红外数据集。数据集包含1个文件夹，“2018AAA0102201-001-北川现场可见光红外数据集”文件夹。“2018AAA0102201-001-北川现场可见光红外数据集”文件夹设置两个子文件夹，文件夹名称为“2018AAA0102201-001-北川现场可见光数据集”文件夹，数据内容为房屋倒塌的废墟周围放置假人和真人，目标为部分遮挡和未遮挡，目标姿态分为行走、蹲坐和站立；碎石废墟周围安排真人移动；树木、桥洞遮挡场景，目标为部分遮挡和未遮挡，分别放置假人和移动的行人进行拍摄的可见光视频帧。容量为4.48G，内含4652帧图片，4652对应标注文本。“2018AAA0102201-001-北川现场红外数据集”文件夹，数据内容为，房屋倒塌的废墟周围放置假人和真人，目标为部分遮挡和未遮挡，目标姿态分为行走、蹲坐和站立；碎石废墟周围安排真人移动；树木、桥洞遮挡场景，目标为部分遮挡和未遮挡，分别放置假人和移动的行人进行拍摄的红外视频帧，容量为0.5G，内含4523帧图片，4523对应标注文本。This dataset comprises visible light and infrared datasets capturing various postures of mannequins and real humans across three scenarios collected at the Beichuan earthquake ruins site. The dataset includes one top-level folder named "2018AAA0102201-001-Beichuan Site Visible Light and Infrared Dataset". This folder contains two subfolders: The first subfolder is "2018AAA0102201-001-Beichuan Site Visible Light Dataset". The data contained herein comprises visible light video frames captured in three types of scenarios: 1) Mannequins and real humans placed around collapsed building ruins, with targets being either partially occluded or unoccluded and in postures of walking, squatting, and standing; 2) Real humans moving around rubble ruins; 3) Scenarios occluded by trees and bridge cavities, where mannequins and moving pedestrians are deployed respectively for image capture, with targets either partially occluded or unoccluded. This subfolder has a storage capacity of 4.48 GB, containing 4652 image frames and 4652 corresponding annotation texts. The second subfolder is "2018AAA0102201-001-Beichuan Site Infrared Dataset". The data contained herein comprises infrared video frames captured in three types of scenarios: 1) Mannequins and real humans placed around collapsed building ruins, with targets being either partially occluded or unoccluded and in postures of walking, squatting, and standing; 2) Real humans moving around rubble ruins; 3) Scenarios occluded by trees and bridge cavities, where mannequins and moving pedestrians are deployed respectively for image capture, with targets either partially occluded or unoccluded. This subfolder has a storage capacity of 0.5 GB, containing 4523 image frames and 4523 corresponding annotation texts.

搜集汇总

背景与挑战

背景概述

该数据集采集自北川地震遗址现场，包含可见光和红外两种模态的数据，主要记录假人和真人在房屋废墟、碎石废墟及树木桥洞等遮挡场景下的不同姿态（如行走、蹲坐、站立）。数据集由可见光部分（4.48GB，4652帧图片）和红外部分（0.5GB，4523帧图片）组成，总数据量为7.57GB，旨在支持模式识别与人工智能相关研究。