Is pointwise convolutin just n_c 1x1 convolutions stacking together?

Hi, just to check if my understanding is correct. I feel like pointwise convolution is just regular 1x1 convolutional layer.

Pointwise convolution is where each 1x1 filter extends across all input channels.
If that’s what you mean by stacking, you’re right.

This is done to change the number of channels in the output, leaving other dimensions match the input dimensions (i.e. height, width).