DHF1K is a video saliency dataset. The dataset consists of 1K high-quality carefully selected video sequences, covering a wide range of scenes, actions, object types, and background complexities.

DHF1K contains a total of 1,000 videos, of which 700 are annotated (600 videos for training and 100 videos for validation) and 300 videos for testing, which will be evaluated on a public server.