Ultra-High-speed Spiking Recognition (UHSR).
Frame cameras suffer from motion blur in high-speed scenarios. Spike cameras have a high temporal resolution (i.e., 50𝜇𝑠) and high dynamic range, which present stronger capability in capturing high-speed motions, but are still hindered by the physical limit in capturing ultra-high speeds under low illumination conditions, as illustrated in the following,
This work proposes a robust representation learning method that utilizes the temporal-spatial contexts of spike streams to address the issues of spiking motion blur, as illustrated in the following,
This work constructs the Ultra-High-speed Spiking Recognition dataset, which contains U-CALTECH and U-CIFAR.
The UHSR dataset is collected using a high-speed motion platform set up in the laboratory under low illumination conditions.
Download: UHSR (6G)
Several samples in the UHSR are visualized in the following,
Zhao J, Zhang S*, Yu Z*, and Huang T. "Recognizing Ultra-High-Speed Moving Objects with Bio-Inspired Spike Camera", in Proceedings of AAAI Conference on Artificial Intelligence. 2024.