Recover depth
The suite has assumed throughout that a photograph holds no depth, because it was lost at capture. That is true of most photographs and not of all of them. A spatial capture holds two views, a Portrait capture holds a measured depth map, and an Android Portrait holds one with its metric range declared. This reads the file to find out which you have, and measures whether the picture can be matched at all — both before any of the estimating starts.
verdict
Where it sits in the file
byte 0
Items
| ID | Type | Size | Bytes | Role |
|---|
Report
What to do with a hit. The suite cannot yet consume a stereo
pair or an auxiliary map directly — that is the recovery tool this page is the front
half of. What it can do today is tell you not to waste the estimating. If a file holds
recorded depth, extract it with the tools listed in the handoff notes, then load the
result into the workbench as a depth map rather than letting the estimate guess.
Measured
Raw measurements
Why the numbers come first. Every figure above is fixed
arithmetic on the pixels, so the same image gives the same answer forever, and you can
copy the JSON and check it. The coach is handed those numbers and never sees the
photograph — it cannot invent a measurement, only interpret one. Sending the picture
to a model and asking what is wrong with it would produce a different confident answer
every time, which is worse than no tool because it feels like measurement.