Grating Cells and Object Recognition
- CABot3 is integrating texture recognition via grating cells.
- They take input from specific orientation line detectors.
- They require input from at least three of these to fire.
- So there are specific orientation and separation grating
cells.
- The grating cells are in a separate subnet. (They cross V1 and
V2 biologically, but Emma Byrne is doing this, so i'ts a good
engineering decision for now.)
- They send activation to the specific regions of the
object recognizers (roughly saying there is an object
here).
- Object Recognition was done in a net called V2 in CABot1 and 2,
but ObjectRec in CABot3.
- In CABot1, position specific recognizers stimulated position general
areas; the general ones were removed for CABot2 and probably 3.
- CABot2 had size (in the visual field) specific receptors.
- One of the tasks was to go to the pyramid, which was met when
it was large in the visual field.
- Note that the CABot1 and CABot2 vision system really only detects
stalactites and pyramids.
- This is all active vision; CAs persist for short periods of time,
but can move as they move through the visual field.
- These are recognized in a postion variant way that is then
fed into position invariant storage.
- Also neurons fatigue, which helps active vision.
