EventCLR: contrastive self-supervised learning for few-shot event-based vision using spiking neural networks

Event-based vision sensors provide significant advantages for high-speed perception, including microsecond temporal resolution, high dynamic range, and low power consumption. When combined with Spiking Neural Networks (SNNs), they are well suited to neuromorphic hardware and energy-efficient applications on embedded systems. However, this potential is severely limited by the scarcity of large-scale labeled datasets required to effectively train such models. In this work, we present EventCLR , a systematic approach to contrastive self-supervised learning for SNNs under label scarcity, a setting that closely matches the practical constraints of most real-world neuromorphic sensing applications. We instantiate this approach with a standard SimCLR-style pipeline adapted to the spiking domain via surrogate gradient training, deliberately keeping the architecture simple, and pair it with a systematic ablation of event-specific augmentation families (spatial, temporal, and polarity transformations). Through extensive experiments on different event datasets (CIFAR10-DVS, N-Caltech101, N-MNIST, DVS-Gesture, and N-Cars), we demonstrate that self-supervised pretraining with subsequent fine-tuning outperforms supervised learning in low-data regimes, achieving consistent gains in few-shot and semi-supervised settings. Our ablation studies show that the combination of the three families works best, and that spatial and temporal augmentations are the most complementary pair. Unlike concurrent work that focuses on large-scale spiking self-supervised pretraining through architectural innovation, EventCLR directly addresses the problem of label efficiency by providing a controlled benchmark for the low-label regime, which remains largely unexplored by existing large-scale approaches. We further show that learned representations transfer across datasets, contributing to efforts for powerful event-based models in label-scarce settings.