Dataset for "An Event-Based Opto-Tactile Skin"

This dataset belongs to the paper “An Event-Based Opto-Tactile Skin”.It includes recordings, processing scripts, and analysis data used to evaluate the performance of a soft, large-area tactile sensing system based on Dynamic Vision Sensors (DVS). The setup uses two DVS cameras in a stereo configuration to detect light changes through a flexible optical waveguide skin. From these events, press locations are estimated using triangulation and DBSCAN clustering. The dataset also contains results related to event rate, latency distribution, and localization accuracy across different pressing trials. Code and processing scripts are available here: https://github.com/event-driven-robotics/optoskin/tree/main Contents of the dataset HDF5 files (.hdf5): grid_path2_v2_camera1.hdf5 – Event recordings from Camera 1 grid_path2_v2_camera2.hdf5 – Event recordings from Camera 2Each file contains event timestamps (ts) and spatial coordinates (x, y) of detected changes. Data is organized as a dictionary with keys for each event attribute. Each HDF5 file contains a dictionary of arrays representing the detected events: Key Description Units ts Event timestamps seconds (s) x X coordinates of events pixels y Y coordinates of events pixels pol Event polarity (+1 for ON, -1 for OFF) unitless CSV files (.csv): snake_grid_press_positions.csv – Recorded positions of tactile presses (in mm), with columns x and y. centroids_camera1_press1_to_250.csv & centroids_camera2_press1_to_250.csv – Computed centroids of activity for presses 1–250 for each camera. Each row corresponds to a press; columns are X and Y coordinates File formats HDF5: Hierarchical format storing arrays for timestamps, x/y positions. Compatible with h5py or bimvee.importHdf5. CSV: Standard comma-separated values with headers for positions and centroids. All units are in millimeters (mm) for positions and seconds (s) for timestamps unless otherwise specified. This dataset is intended for researchers in event-driven perception, robotics, and tactile sensing to reproduce the experiments in the paper or to benchmark new algorithms.