Fig. 3From: Three-dimensional convolutional restricted Boltzmann machine for human behavior recognition from RGB-D videoa 2D convolution and b spatial temporal convolutionBack to article page