objrec

README

Bench Objrec

bench_objrec is supposed to be an easy-to-understand, easy-to-run network that nevertheless has the same performance characteristics as the big LVis network. As compared to examples/bench_lvis, bench_objrec is more difficult to understand (uses Looper, lots of logging code), but actually converges and has a working GUI.

This network is similar to lvis/sims/objrec. See the README over there for more detailed explanations.

Changes:

• Works with most recent Axon version.
• Prints a few basic stats to stdout

TODOs:

• Delete lots of the unnecessary code (like everything in logging, except the most basic stats e.g. CorSim)
• Instead of being package main, turn the main() into a BenchmarkObjrec(b *testing.B) to make it easier to run & profile. Currently, this prohibits the use of a GUI (the window never gets created, just blocks here forever).

Documentation

Overview

objrec explores how a hierarchy of areas in the ventral stream of visual processing (up to inferotemporal (IT) cortex) can produce robust object recognition that is invariant to changes in position, size, etc of retinal input images.