Spatial pooling is a pooling operation that summarizes local spatial regions in feature maps. It is commonly used in convolutional neural networks to reduce spatial resolution, improve robustness, and preserve salient features.
See pooling.