|
|
|
|
|
by johndough
1402 days ago
|
|
The problem is not AI - many of the best "AI" publications are open source after all. The problem is that hardware vendors do not provide access to their signal processing chip's internals. It is damn near impossible to even make a phone call these days without some obscure binary blob or magic chips that nobody knows what they do but are able to control every aspect of a phone. |
|
In my experience there's also very little detailed documentation about what cutting-edge phone cameras are doing.
You can get some vague descriptions (focus stacking? exposure stacking? ISO stacking? ML bokeh? Special handling of faces in multiracial groups? Shake compensation? Super-resolution?) which is all very well shooting from a tripod - yet modern phone cameras do their magic at 4k 60fps even while moving? All while running on battery?