Mixtac: A Bio-inspired Hybrid Tactile Sensor Dataset with Synergistic Event-Frame Perception

This dataset contains raw data collected for the evaluation and training of the Mixtac tactile sensor and its associated deep learning model, FGER-Net. Mixtac is a novel bio-inspired hybrid tactile sensor that integrates "slow-adapting" (image frames) and "fast-adapting" (event streams) channels to resolve the trade-off between dynamic response and long-term static stability in vision-based tactile sensing. The data was acquired using a DAVIS346 event camera and includes synchronized grayscale frames, asynchronous event streams, and ground truth force data from a 6-axis force sensor. The dataset covers various experimental scenarios, including high-frequency vibration sensing (50-250 Hz), dynamic force tracking, long-term static force estimation, hybrid transient impact and static hold tests, and extended force range (0-12 N) validation. This dataset is valuable for research in visual-tactile perception, robotic dexterous manipulation, and multi-modal sensor fusion.