iLab Neuromorphic Vision Toolkit

Each call takes a reference to an appropriate ModelComponent derivative object, which must furthermore be the appropriate specific derived type for the module type.

Each call takes a boolean parameter 'makeSubComp'. If true, the object passed as argument will become a subcomponent of us, thus you will not need to separately register it with ModelManager. Yields a fatal error if we already have one module of that type (name). Typically, you want to set this flag to true in models where the Brain is your top-level (root) component; in models that contains objects for whom the Brain is a subcomponent, typically they will also have our modules as subcomponents, so we should not set this flag to true

HISTORICAL NOTE: For now, these replace the old addModule() function -- the reason was that everybody who called addModule() was doing it with a type known at compile time (i.e., they called it with a literal "VisualCortex" string and a nub::soft_ref<VisualCortex>), so it's more straightforward to just have separate functions for each type. This approach also avoids a round of up+down casting (with addModule, you might have a VisualCortex object, which then got casted to ModuleComponent for the addModule() call, only to be dynamic-casted back to VisualCortex inside addModule()). If, in the future, we end up with a system in which Brain doesn't need to know the actual types of all its modules (e.g., it just has a bag of BrainModule objects), then it could make sense to go back to the addModule() system.


Public Member Functions
nub::ref< PrefrontalCortex >	getPFC () const
	Get the PFC.
nub::ref< Retina >	getRET () const
	Get the Retina.
nub::ref< VisualCortex >	getVC () const
	Get the VisualCortex.
nub::ref< EyeHeadController >	getEHC () const
	Get the EyeHeadController.
nub::ref< SimulationViewer >	getSV () const
	Get the SimulationViewer.
nub::ref< ShapeEstimator >	getSE () const
	Get the ShapeEstimator.
nub::ref< SaliencyMap >	getSM () const
	Get the SaliencyMap.
nub::ref< TaskRelevanceMap >	getTRM () const
	Get the TaskRelevanceMap.
nub::ref< AttentionGuidanceMap >	getAGM () const
	Get the AttentionGuidanceMap.
nub::ref< AttentionGate >	getAG () const
	Get the AttentionGate.
nub::ref< WinnerTakeAll >	getWTA () const
	Get the WinnerTakeAll.
nub::ref< InferoTemporal >	getIT () const
	Get the InferoTemporal.
nub::ref< GistEstimator >	getGE () const
	Get the GistEstimator.
nub::ref< TargetChecker >	getTC () const
	Get the TargetChecker.
nub::ref< RetinaToWorld >	getRTW () const
	Get the RetinaToWorld.
void	drawResults (Image< PixRGB< byte > > &traj, const Rectangle &foa)
	Pass along to our subcomponents (VisualCortex in particular).
Constructors, destructors, and run-time "modules"
Typically, you instantiate a IRoomBrain and then add various plug-in "modules" to it (using setRET(), setVC(), etc.). If you decide so, each module you add may become a subcomponent of your IRoomBrain object; if so, you should not do a ModelManager::addSubComponent() on them, but just do it once for all on your IRoomBrain object instead.
	Brain (OptionManager &mgr, const std::string &descrName="Brain", const std::string &tagName="Brain")
	Constructor.
virtual	~Brain ()
	Virtual destructor for safe inheritance.
Module setters
Each call takes a reference to an appropriate ModelComponent derivative object, which must furthermore be the appropriate specific derived type for the module type. Each call takes a boolean parameter 'makeSubComp'. If true, the object passed as argument will become a subcomponent of us, thus you will not need to separately register it with ModelManager. Yields a fatal error if we already have one module of that type (name). Typically, you want to set this flag to true in models where the Brain is your top-level (root) component; in models that contains objects for whom the Brain is a subcomponent, typically they will also have our modules as subcomponents, so we should not set this flag to true HISTORICAL NOTE: For now, these replace the old addModule() function -- the reason was that everybody who called addModule() was doing it with a type known at compile time (i.e., they called it with a literal "VisualCortex" string and a nub::soft_ref<VisualCortex>), so it's more straightforward to just have separate functions for each type. This approach also avoids a round of up+down casting (with addModule, you might have a VisualCortex object, which then got casted to ModuleComponent for the addModule() call, only to be dynamic-casted back to VisualCortex inside addModule()). If, in the future, we end up with a system in which Brain doesn't need to know the actual types of all its modules (e.g., it just has a bag of BrainModule objects), then it could make sense to go back to the addModule() system.
void	setPFC (nub::ref< PrefrontalCortex > mod, bool makeSubComp=false)
	Set the PrefrontalCortex.
void	setRET (nub::ref< Retina > mod, bool makeSubComp=false)
	Set the Retina.
void	setVC (nub::ref< VisualCortex > mod, bool makeSubComp=false)
	Set the VisualCortex.
void	setEHC (nub::ref< EyeHeadController > mod, bool makeSubComp=false)
	Set the EyeHeadController.
void	setSV (nub::ref< SimulationViewer > mod, bool makeSubComp=false)
	Set the SimulationViewer.
void	setSE (nub::ref< ShapeEstimator > mod, bool makeSubComp=false)
	Set the ShapeEstimator.
void	setSM (nub::ref< SaliencyMap > mod, bool makeSubComp=false)
	Set the SaliencyMap.
void	setTRM (nub::ref< TaskRelevanceMap > mod, bool makeSubComp=false)
	Set the TaskRelevanceMap.
void	setAGM (nub::ref< AttentionGuidanceMap > mod, bool makeSubComp=false)
	Set the AttentionGuidanceMap.
void	setAG (nub::ref< AttentionGate > mod, bool makeSubComp=false)
	Set the AttentionGate.
void	setWTA (nub::ref< WinnerTakeAll > mod, bool makeSubComp=false)
	Set the WinnerTakeAll.
void	setIT (nub::ref< InferoTemporal > mod, bool makeSubComp=false)
	Set the InferoTemporal.
void	setGE (nub::ref< GistEstimator > mod, bool makeSubComp=false)
	Set the GistEstimator.
void	setTC (nub::ref< TargetChecker > mod, bool makeSubComp=false)
	Set the TargetChecker.
void	setRTW (nub::ref< RetinaToWorld > mod, bool makeSubComp=false)
	Set the RetinaToWorld.
Brain simulation functions
void	setTargetMask (const Image< byte > &targets)
	Specify a target mask.
void	input (const Image< PixRGB< byte > > &inp, nub::soft_ref< SimEventQueue > q)
	Load a new frame and pass it through visual cortex.
virtual void	evolve (SimEventQueue &q)
	Evolve all of our modules.
SimStatus	evolve (bool useLowMem)
	Evolve all of our modules.
Protected Member Functions
virtual void	start1 ()
	do some additional config at start time; see ModelComponent.H
Classes
struct	Impl
	Private implementation struct for Brain. More...

Brain Class Reference

Detailed Description

Public Member Functions

Protected Member Functions

Classes

Constructor & Destructor Documentation

Member Function Documentation