MADSNet: A Multi-Scale Axial-Attentive Dense Spiking Network for Spike Camera Image Reconstruction

A new arXiv paper (2611.12980v1) introduces MADSNet, a Multi-scale Axial-attentive Dense Spiking Network for reconstructing images from spike camera data. On the SREDS benchmark, MADSNet outperforms SSIR by 4.3 dB PSNR and Spk2ImgNet by 1.94 dB, while achieving roughly 15× lower latency and using about 1/40 of the estimated energy compared with Spk2ImgNet. The authors state the results show carefully designed spiking neural network architectures can improve reconstruction fidelity while preserving computational efficiency.

arXiv:2611.12980v1 Announce Type: new Abstract: Spike cameras encode dynamic scenes into high-temporal-resolution binary spike streams for efficient high-speed vision. Although spiking neural networks SNNs can process such signals with sparse and accumulate-only computation, existing SNN-based reconstruction methods are typically shallow and still lag behind ANN-based models in reconstruction fidelity. This paper proposes MADSNet, a Multi-scale Axial-attentive Dense Spiking Network for spike camera image reconstruction. MADSNet adopts a hierarchical spiking encoder--decoder, where a multi-scale spike stem extracts early features, axial spike attention captures long-range dependencies, a dense spiking bottleneck enhances feature reuse, and a spike-to-continuous decoder reconstructs continuous intensity images from spike-domain representations. Experiments on gray-scale and color spike reconstruction show that MADSNet outperforms SSIR by 4.3 dB PSNR and Spk2ImgNet by 1.94 dB on SREDS. Compared with Spk2ImgNet, it achieves about 15$\times$ lower latency and uses only about 1/40 of the estimated energy, demonstrating that carefully designed SNN architectures can improve reconstruction quality while preserving computational efficiency.