ISFire-4
收藏资源简介:
ISFire-4 是一个用于室内火灾和烟雾检测的多类别目标检测数据集,包含边界框标注,格式为 YOLO。该数据集在论文《FireNet-D5: A Genetic Algorithm-Optimized Architecture with BiFPN and SE Attention for Multi-Class Indoor Fire and Smoke Detection》中首次提出。数据集包含4个类别:黑烟(black smoke)、蓝火(blue fire)、白灰烟(white-gray smoke)和黄橙火(yellow-orange fire)。总计7,370张图像,共11,010个实例。数据以压缩包 dataset.zip 提供,内部划分为 train、valid 和 test 三个子集,每个子集下分别包含 images(图像)和 labels(标签)文件夹。标签文件采用 YOLO 格式,每行记录为:class_id x_center y_center width height,所有坐标值均归一化到 [0,1]。数据集采用 CC-BY-4.0 许可证。
ISFire-4 is a multi-class object detection dataset for indoor fire and smoke detection, with bounding box annotations in YOLO format. It was first introduced in the paper FireNet-D5: A Genetic Algorithm-Optimized Architecture with BiFPN and SE Attention for Multi-Class Indoor Fire and Smoke Detection. The dataset contains 4 classes: black smoke, blue fire, white-gray smoke, and yellow-orange fire. There are a total of 7,370 images with 11,010 instances. The data is provided as a compressed package dataset.zip, internally divided into train, valid, and test subsets, each containing images and labels folders. The label files are in YOLO format, with each line recording: class_id x_center y_center width height, all coordinates normalized to [0,1]. The dataset is licensed under CC-BY-4.0.
ISFire-4: 室内火焰与烟雾检测数据集
该数据集专注于室内场景下的火焰与烟雾多类别检测任务,提供了基于YOLO格式的边界框标注,适用于目标检测模型的训练与评估。
- 任务类型:目标检测(Object Detection)
- 许可证:CC-BY-4.0
类别信息
数据集包含4个类别,具体如下:
- 黑烟(black smoke)
- 蓝色火焰(blue fire)
- 白灰色烟雾(white-gray smoke)
- 黄橙色火焰(yellow-orange fire)
数据规模
- 图像总数:7,370张
- 标注实例总数:11,010个
数据组织结构
数据集以压缩包 dataset.zip 形式提供,解压后包含三个子目录:train/、valid/、test/。每个子目录下分别存放 images/(图像文件)和 labels/(标注文件)。标注文件采用YOLO格式,每行记录内容为:class_id x_center y_center width height,其中坐标值均经过归一化处理。
来源与引用
该数据集源自论文 FireNet-D5: A Genetic Algorithm-Optimized Architecture with BiFPN and SE Attention for Multi-Class Indoor Fire and Smoke Detection,如使用本数据集,请引用该论文。




