PEER Hub ImageNet (Φ-Net)-Detection Extention

This dataset is an extension of tasks 7 and 8 of the PEER Hub ImageNet (Φ-Net) classification dataset. The data provided here is annotated for the detection tasks in structural health monitoring. The dataset comprises two main related tasks (damage type and damage level) or sub-datasets. The damage type has four class instances: (1) Undamaged, (2) Flexural, (3) Shear, and (4) Combined Damage. The damage level also has four class instances: (1) Undamaged, (2) Minor, (3) Moderate, and (4) Heavy Damage. Further, data augmentation techniques including Mosiac, CutMix, and MixUp have been applied to increase the dataset size from 4585 to 11005 for the damage type and from 4636 to 10904 for the damage level.本数据集为PEER Hub ImageNet（Φ-Net）分类数据集任务7与任务8的扩展版本。本次提供的数据针对结构健康监测领域的检测任务完成了标注。该数据集包含两项核心关联任务（损伤类型与损伤等级）及其对应的子数据集。其中损伤类型共设4个类别样本：(1) 无损伤，(2) 弯曲损伤，(3) 剪切损伤，(4) 复合损伤；损伤等级同样包含4个类别样本：(1) 无损伤，(2) 轻微损伤，(3) 中度损伤，(4) 重度损伤。此外，本数据集应用了Mosiac、CutMix与MixUp等数据增强技术，将损伤类型子数据集的规模从4585扩充至11005，损伤等级子数据集的规模从4636扩充至10904。

搜集汇总

背景与挑战

背景概述

该数据集是PEER Hub ImageNet (Φ-Net)的检测扩展，专注于结构健康监测，包含损伤类型和损伤级别两类任务，每类有四个实例。通过数据增强技术，数据集规模从约4500增至约11000，适用于土木和建筑工程的检测任务。