How to get an understanding of the in- and output shapes?

[JohannesBauer97]

I'm searching for the right docs/code to understand the input and outputs of the models.
For example, the RetinaNet model for object detection, after converting it to my backend CoreML using mmdeploy, I can see it has the following input and output shapes:

• Input: (1, 3, 800, 1344)
• Output: Float32 & Int32 (one dimension for label and dets)

I guess the in- and outputs aren't changed when converting to different models, so CoreML should be just an example.

For the input shape, the 3 is probably the color channels, 800 and 1344 height and width. But how can I find out what's the first (1) dimension for?

Also for the output shape dets I would have expected x, y, w, h and not just one dimension.

tl;dr: Where to find a description of the input and output shapes for the detection models?