My understanding of S3 is
You don't want your client hit your server then hit s3 via urls.py. Your client is better off getting files(images) from s3 directly.
Therefore you will want to return the whole url(image_field.url) rather than relative path(image_field.name or similar) to your client to use.
But then, what if you move from S3 to some other service?
Returning whole URL to client would be an OK practice assuming you might want to switch from amazon S3 to something else later?