[Deep Learning] I have a question about the structure of Google Net's Inception.

I'm asking you this question because I suddenly got suspicious while reading the Google Net v1 thesis review.

Google Net undergoes convolution of several sizes through Inception.

When a feature map goes through several sizes of convolution

The final feature map will also have different sizes

The feature maps are combined into one at the end of Inception to form a channel

Is there nothing wrong with configuring feature maps of different sizes in this way?

Or is there a process I don't know?