Semantic Segmentation Image Dataset for Balconies Facade and Hanging Objects Safety Risks

In the context of rapid urban development, the facades and hanging objects on balconies of apartments and homes have increasingly become part of safety concerns, posing potential threats to safety in daily life. Existing manual safety inspections are not only time-consuming and prone to omissions but also unable to effectively support real-time monitoring needs. This dataset aims to improve the efficiency of identifying safety hazards on balconies and hanging objects through precise semantic segmentation, meeting intelligent monitoring business needs. Image data of balconies in urban and rural residential buildings are collected by drone equipment from different heights and angles to ensure diversity and authenticity. During the data collection process, quality is strictly controlled using multiple annotation rounds and consistency checks, with all annotation results audited by a 30-member team with architectural and safety expertise. Data preprocessing adopts advanced technologies such as denoising, cropping, and enhancement, stored in JPG format and organized by building type and region. The annotation accuracy of this dataset reaches 95%, ensuring high consistency and completeness, utilizing improved annotation techniques and quality evaluation methods to enhance monitoring efficiency and safety. Compared to similar datasets, this dataset has unique advantages in terms of data breadth and rarity, providing diverse scenarios of urban and rural areas, and supporting multi-task extension applications. Through the application of this dataset, the accuracy of risk detection for balconies façades and their hanging objects has increased by 20%, showing significant application value.