Start with the why:
The additional depth modes available in our framework are useful in certain scenarios, including:
- Supporting closer-in minimum depth (by a factor of 2): Extended disparity
- Eliminating shadowing from running disparity in only one direction: left/right check (runs disparity in both directions)
- Allowing further maximum distance, and better accuracy w/in given range: subpixel support
Move to the how:
Analyze the memory/resource use of enabling these modes through the API, and implement such that limits are placed on other resources appropriately (e.g. fewer SHAVES available for neural compute).
Move to the what:
Support the following additional depth modes:
Start with the
why:The additional depth modes available in our framework are useful in certain scenarios, including:
Move to the
how:Analyze the memory/resource use of enabling these modes through the API, and implement such that limits are placed on other resources appropriately (e.g. fewer SHAVES available for neural compute).
Move to the
what:Support the following additional depth modes: