SE-SNN: Squeeze-and-Excitation Enhanced Spiking Neural Networks with Learnable Neuron Dynamics for Event-Based Vision

Spiking Neural Networks (SNNs) have emerged as a promising paradigm for energy-efficient neuromorphic computing, particularly when processing asynchronous event streams from dynamic vision sensors (DVS). However, SNNs often suffer from limited representational capacity and suboptimal feature recalibration compared to their artificial counterparts. To address these challenges, we propose SE-SNN, a novel architecture that integrates Squeeze-and-Excitation (SE) blocks into deep residual SNNs, enabling channel-wise attention without spike generation in the gating mechanism. Furthermore, we introduce a Robust Parametric Leaky Integrate-and-Fire (RobustPLIF) neuron model with learnable membrane time constant (τ) and firing threshold (Vth), allowing adaptive temporal dynamics per layer. Our model is trained on the CIFAR10-DVS dataset.Experimental results demonstrate that SE-SNN achieves state-of-the-art accuracy of 78.8 % on CIFAR10-DVS with only 16 time steps, significantly outperforming baseline SNNs while maintaining biological plausibility and hardware efficiency. Ablation studies confirm the individual contributions of SE blocks and learnable neuron parameters to performance gains.