But TENNs go further and start treating streaming inputs like video like a time series of frames, performing a 3D convolution comprising of a temporal convolution on the time axis and a spatial ...