Motor Speed Derating Dataset

This dataset supports the development and evaluation of a vision-based motor-speed derating pipeline for safe human–machine interaction. It contains 10 recorded episodes of a hand approaching a monitored motor-control setup, comprising 5,493 RGB frames acquired at 800 × 600 pixels and approximately 30 frames per second. Each frame is associated with a manually annotated normalized motor-speed target in the range [0, 1], where 0 and 1 represent the application-defined minimum and maximum speed levels. The dataset also provides control-rate labels sampled at 75 Hz. These labels were obtained by linearly interpolating the frame-level annotations and applying temporal smoothing.The archive includes the source videos, extracted RGB frames, episode metadata, the frame-level manifest (data.csv), and the 75 Hz control labels (controls_75hz.csv). The dataset was collected in a controlled laboratory setup and is intended for research and reproducibility purposes. Models trained on it may not generalize to different camera viewpoints, illumination conditions, backgrounds, hand poses, or motor-control hardware arrangements.