pyronear/pyro-sdis

Pyro-SDIS数据集是一个用于野火烟雾检测的AI模型训练数据集，由法国消防和救援服务（SDIS）与Pyronear协会合作开发。数据集包含33,636张图像，其中28,103张图像包含烟雾，31,975个烟雾实例。每张图像都附有YOLO格式的注释，适用于Ultralytics YOLO框架。数据集的结构包括图像路径、注释、元数据（合作伙伴、相机标识、拍摄日期、图像名称）和数据集划分（训练集或验证集）。数据集分为训练集和验证集，训练集包含29,537张图像，验证集包含4,099张图像。README还提供了如何将数据集导出为Ultralytics格式的详细步骤，以及如何使用Ultralytics YOLO框架进行训练的说明。

The Pyro-SDIS dataset is designed for wildfire smoke detection using AI models, developed in collaboration with the Fire and Rescue Services (SDIS) in France and the Pyronear association. The dataset contains 33,636 images, including 28,103 images with smoke and 31,975 smoke instances. Each image is accompanied by YOLO-format annotations, making it compatible with the Ultralytics YOLO framework. The dataset structure includes image paths, annotations, metadata (partner, camera identifier, capture date, image name), and dataset splits (train or val). The dataset is divided into a training set with 29,537 images and a validation set with 4,099 images. The README also provides detailed steps on how to export the dataset into Ultralytics format and instructions on training using the Ultralytics YOLO framework.

搜集汇总

背景与挑战

背景概述

Pyro-SDIS数据集包含33,636张图像，其中28,103张带有烟雾标注，共计31,975个烟雾实例。数据集支持Ultralytics YOLO框架，适用于野火烟雾检测模型的训练，预计2025年1月将发布包含更多图像和精细标注的完整版本。