AI-Based Classification of Mosquito-Prone and Non-Prone Urban Environments Using Image Data in Bangladesh

This dataset contains approximately 5,000 ground-level images collected from more than 25 urban locations across Dhaka, Jessore, and Magura in Bangladesh to support AI-based mosquito breeding risk classification. The images represent diverse urban environments captured under varied weather, lighting, and seasonal conditions during pre-monsoon and early-monsoon periods. The dataset is organized into two folders corresponding to the risk classes: Prone, Non prone where each folder contains images of that category. All images were manually labeled by three team members independently, with a consensus review achieving Cohen's kappa of 0.87, indicating very high inter-annotator agreement. Images were captured using consumer-grade smartphones and preprocessed to a standardized 640 × 480-pixel resolution in 4:3 aspect ratio. The dataset is intended for image classification tasks to train, validate, and evaluate models that can automatically detect and classify urban environments by mosquito breeding risk level. Categories for this data: Prone (visible stagnant water, garbage accumulation, clogged drains, partially wet surfaces with potential breeding conditions), Non prone (clean dry roads and well-maintained open spaces)本数据集包含约5000张地面视角图像，采集自孟加拉国达卡、杰索尔与马古拉三地的25处以上城区点位，旨在支撑基于人工智能（AI）的蚊虫滋生风险分类任务。 所采集图像涵盖多样化城区场景，拍摄于季风前及季风早期阶段，涵盖不同天气、光照与季节条件。 本数据集按照风险类别划分为两个文件夹：易滋生（Prone）与非易滋生（Non prone），每个文件夹存储对应类别的图像。 所有图像均由三名团队成员独立完成手动标注，经一致性审核后，标注者间的科恩kappa系数（Cohen's kappa）达0.87，表明标注一致性极高。 图像均采用消费级智能手机拍摄，并预处理至统一的640×480像素分辨率，宽高比为4:3。 本数据集适用于图像分类任务，可用于训练、验证与评估可自动检测城区场景并按蚊虫滋生风险等级进行分类的模型。 本数据集的类别说明如下： 易滋生（Prone）：存在可见积水、垃圾堆积、排水管道堵塞及局部潮湿表面等潜在滋生条件的场景； 非易滋生（Non prone）：路面整洁干燥、户外空间维护良好的场景。