Sudden Fall Action Recognition Dataset

The current security industry faces an increasing number of fall incidents involving the elderly, especially in public places such as hospitals and stations. These incidents not only affect the safety of the elderly but also impose a huge burden on society. Existing surveillance systems often cannot promptly identify such accidental incidents, causing delays in assistance. This dataset aims to accurately recognize the actions of elderly people suddenly falling under non-exercise conditions through video analysis technology to improve emergency response capabilities. Data collection uses high-resolution surveillance cameras in real environments such as elderly communities, hospitals, etc. We ensure high data quality through multiple rounds of annotation and expert review. Meanwhile, the data storage format is MP4, organized by video ID for convenient subsequent analysis and processing.