Convolution
Image will be of shape: (H * W * channels)
Our kernel will also be of shape (Hk * Wk * channels)
The number of channels is same as it is volume convolution.
In case of 1 * 1 convolution, the convolution helps decrease or increase the feature map dimensions.