PUCK: Parallel Surface and Convolution-kernel Tracking for Event-Based Cameras

The dataset can be used to test your event-based tracking algorithm. In the experiment, a static and moving event-camera (640x480 pixels) mounted on the iCub robot head recorded different sequences of an air hockey puck thrown by a human player. The dataset includes event-driven data and ground truth of 10 sequences for the static camera and 10 sequences for the moving camera, 30 s each. To import .log files containing events, we suggest bimvee Python library. Specifically, use the functions to import .log files: data = importIitYarp(filePathOrName=input_path) Ground-truth .csv files have 5 columns, each one corresponding to a different measure: timestamp | roi x min | roi y min | roi x max | roi y max The roi is the puck bounding box. From the roi columns, the puck center position can be computed.