It means that instead of (only) doing convolution in spatial dimensions, it also(/instead) happens in the temporal dimension. A good resource for the "instead"

It means that instead of (only) doing convolution in spatial dimensions, it also(/instead) happens in the temporal dimension. A good resource for the "instead" case: https://unit8.com/resources/temporal-convolutional-networks-and-forecasting/ https://unit8.com/resources/temporal-convolutional-networks-... The "also" case is an example of 3D convolution, an example of a paper that uses it: https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Tran_Learning_Spatiotemporal_Features_ICCV_2015_paper.pdf https://www.cv-foundation.org/openaccess/content_iccv_2015/p...