This is a great question and something I realized I have somewhat glanced over on this blog. It's important to note that you will most likely NOT see shapes when just getting started. If you are sighted, like this person, or blind and only use echolocation as a guideline for determining your generic environment, you have got a ways to go before you will actually be able to distinguish shapes.
Echolocation should be trained one little step at a time. If you can sense distance or at least have a feeling whether or not something is in front of you, the next step is "edge finding". Locate the object until it drops off. There's the edge! Keep training that however, because you'll notice that the edges come into focus better and better over time. Move the position of your head back and forth over the object edge and really try to identify where the exact edge is. It's good to do edge finding on hard, flat objects with crisp lines and nothing in the background such that the signal is completely different on either side of the edge. A doorjamb is a good tool for edge finding.
Eventually, with this progressive practice, you may be able to see entire shapes at a glance, but I think that even very adept, blind from birth, echolocators still sometimes struggle with seeing full shapes. Instead they might see edges, one at a time, and have to piece them together over the course of a few clicks.