* This dataset supports online use.Click here to jump.

The FADE (FAlling Object DEtection around Buildings) dataset is a large and diverse video dataset specially built for detecting falling objects around buildings. The dataset was jointly developed by Zhigang Tu, Zitao Gao, Zhengbo Zhang, Chunluan Zhou, Junsong Yuan, and Bo Du in 2024. The related paper results are "FADE: A Dataset for Detecting Falling Objects around Buildings in Video".

The FADE dataset contains 1,881 videos covering 18 scenes, 8 different categories of falling objects, 4 different weather conditions, and 4 video resolutions. In addition, the researchers developed a new object detection method, FADE-Net, which effectively utilizes motion information to generate small-sized but high-quality detection proposals to improve the accuracy of detecting falling objects. By comparing with existing general object detection, video object detection, and moving object detection methods, FADE-Net shows significant performance advantages on the FADE dataset, providing an effective benchmark for future related research.

The FADE dataset was created to address the problem of falling objects from high floors that can cause serious harm to pedestrians. Since falling objects in surveillance videos can be difficult to capture manually due to their small size, fast movement, and complex background, it is particularly important to develop automated detection methods. The diversity and professionalism of the FADE dataset make it a valuable resource for studying falling object detection around buildings.