Fig. 2: The ESOT500 dataset for high-dynamic event-driven perception.
From: Bridging the latency gap with a continuous stream evaluation framework in event-driven perception
a Representative event stream samples from the low-resolution ESOT500-L subset (346 × 260, top row) and high-resolution ESOT500-H subset (1280 × 720, bottom row), showcasing diverse high-dynamic scenarios (e.g., flag waving, bicycle riding, car driving, cap shaking, football playing, monkey swinging, fan rotating, pigeon taking off, bottle spinning, breakdancing). b Scene category distribution for ESOT500-L (left) and ESOT500-H (right), quantifying the percentage of sequences across attributes like object similarity, background complexity, deformation, occlusion, motion speed, and indoor/outdoor setting. c Comparison of event-based object tracking datasets. ESOT500-L and ESOT500-H stand out with 500 Hz time-aligned annotations, high resolution (up to 1280 × 720 for ESOT500-H), and diverse scene coverage, addressing gaps in prior datasets (e.g., low annotation frequency, lack of time-aligned labels)5,21,24,63,71,72,73,74,75,76,77,78,79,80,81,82,83,84.
