Unsupervised spiking convolutional neural networks (SCNNs) exhibit improved performance to object-optimized CNNs in predicting fMRI responses induced by visual stimuli and achieve promising results on image identification and reconstruction tasks.