Event based Drone Detection and Segmentation (EDDS)

Drone Event Dataset This dataset contains 15 event recordings (275s total) of a flying drone with event segmentation and 100Hz bounding box annotations.7 hand-held recordings, 8 stationary recordings.`motion01` - `motion04` and `stationary01` - `stationary04` were recorded inside and contain walls, windows, furniture and a patterned floor.`motion05` - `motion07` and `stationary05` - `stationary08` were recorded outside and contain grass, flower beds, brush, building walls and the cloudy sky. Camera used: Prophesee EVK4 Files *.raw files contain the event recordings in Prophesee EVT RAW format. *.mp4 files contain false frame videos at 100Hz frame rate as generated by OpenEB. *.csv files contain the bounding box annotations in CSV format with columns (`frame_idx`, `left`, `width`, `top`, `height`).`frame_idx` corresponds to the frame index of the 100Hz mp4 file.This means, `frame_idx` $i$ corresponds to the time interval $[i * 10ms, (i+1) * 10ms)$ in the event recording. motion??/trajectory*.csv and stationary??/trajectory*.csv files contain the trajectory predictions in CSV format with one file per continuous trajectory segment containing one row per time step with columns (`x`, `y`, `t`) with units (px, px, us). motion??/segmentation*.csv and stationary??/segmentation*.csv files contain the event segmentations in CSV format with one file per continuous trajectory segment containing one row per event with columns (`x`, `y`, `t`, `p`) with units (px, px, us, bool).