Spike Stream

Spike streams, generated by high-speed spike cameras, represent visual information as sequences of binary events, offering advantages in temporal resolution and dynamic range over traditional frame-based cameras. Current research focuses on developing algorithms and models, such as recurrent spiking transformers and variations of neural radiance fields (NeRFs), to reconstruct images and 3D scenes from these noisy, temporally rich data streams, addressing challenges like motion blur and low-light conditions. This work is significant for advancing high-speed imaging applications, including autonomous driving and virtual reality, by enabling accurate 3D reconstruction and novel view synthesis from challenging visual inputs. Improved simulation tools are also contributing to the development of more robust and accurate spike-based vision systems.