Prompt-to-Gesture: Measuring the Capabilities of Image-to-Video Deictic Gesture Generation

Background This dataset is a computer vision and video dataset built for training and evaluating downstream deictic (pointing) gesture recognition models in human-robot interaction (HRI) settings. Traditional gesture data collection requires expensive human recording sessions, which severely limits data scale and diversity. This dataset addresses that bottleneck by mixing real human data with realistic, image-to-video generated synthetic videos to train more robust models.… See the full description on the dataset page:

Source: Hugging Face Hub (sano90/prompt-to-gesture). Metadata imported from the dataset’s Hub tags.