DataProcessor executing a Kinesis::Vision pipeline on pixel container data. More...

#include <VisionProcessor.hpp>

Inheritance diagram for MayaFlux::Kakshya::VisionProcessor:

Collaboration diagram for MayaFlux::Kakshya::VisionProcessor:

Public Member Functions
const Kinesis::Vision::VisionResult &	get_result () const
	The result of the last successful process() call.

std::shared_ptr< Vruta::BroadcastSource< Kinesis::Vision::VisionResult > >	get_result_source ()
	Shared BroadcastSource signalled with each VisionResult after a successful process() call.

bool	is_processing () const override
	Checks if the processor is currently performing processing.

void	on_attach (const std::shared_ptr< SignalSourceContainer > &container) override
	Cache frame geometry from get_structure().

void	on_detach (const std::shared_ptr< SignalSourceContainer > &container) override
	Clear cached geometry and reset executor state.

void	process (const std::shared_ptr< SignalSourceContainer > &container) override
	Execute the VisionSequence on processed_data[0].

void	set_sequence (Kinesis::Vision::VisionSequence sequence)
	Replace the pipeline and reset inter-frame executor state.

	VisionProcessor (Kinesis::Vision::VisionSequence sequence, bool force_cpu=false)
	Construct with the vision pipeline to execute each process() call.

	~VisionProcessor () override=default

Public Member Functions inherited from MayaFlux::Kakshya::DataProcessor
virtual	~DataProcessor ()=default
	Virtual destructor for proper cleanup.

Private Attributes
Kinesis::Vision::VisionExecutor	m_cpu_executor

std::unique_ptr< Yantra::VisionGpuExecutor >	m_executor

std::vector< float >	m_float_storage

bool	m_force_cpu {}

std::shared_ptr< Core::VKImage >	m_gpu_frame

uint32_t	m_height {}

std::atomic< bool >	m_is_processing { false }

Kinesis::Vision::VisionResult	m_result

std::shared_ptr< Vruta::BroadcastSource< Kinesis::Vision::VisionResult > >	m_result_source

Kinesis::Vision::VisionSequence	m_sequence

std::shared_ptr< Buffers::VKBuffer >	m_upload_staging

uint32_t	m_width {}

Detailed Description

DataProcessor executing a Kinesis::Vision pipeline on pixel container data.

Compatible with three container types, each accessed via its native normalised float path:

VideoStreamContainer: processed_frame_as_float(0) on processed_data
WindowContainer: processed_frame_as_float(0) on processed_data
TextureContainer: as_normalised_float(0) on m_data

on_attach throws std::invalid_argument for any other container type. Runs the configured VisionSequence through VisionExecutor and stores the VisionResult as member state for polling via get_result().

Width and height are read from get_structure() at on_attach time. m_float_storage is reused across calls to avoid per-frame allocation.

See also: FrameAccessProcessor, WindowAccessProcessor, VisionExecutor, VisionSequence

Definition at line 33 of file VisionProcessor.hpp.

The documentation for this class was generated from the following files:

Public Member Functions

Private Attributes

Detailed Description