select_label() [2/2]

std::shared_ptr< Core::VKImage > MayaFlux::Yantra::VisionGpuExecutor::select_label	(	VisionGpuContexts &	contexts,
		const std::shared_ptr< Core::VKImage > &	source,
		uint32_t	target_label,
		uint32_t	w,
		uint32_t	h
	)

Isolate one ConnectedComponents label's silhouette on GPU, through the explicit context set that produced the label buffer.

Dispatches vision_label_select.comp through contexts.component_contours, the same context ConnectedComponents/FindContours already own: reads dense_label in place at its existing binding, no cross-context buffer copy. Requires a ConnectedComponents step to have already run against these exact contexts (this run or an earlier one); dense_label holds whichever frame's labels were computed last.

Parameters

contexts	The same context set a prior run() populated.
source	Frame to select pixels from, eShaderReadOnlyOptimal.
target_label	1-based label id, matching BoundingBox::label_id / Contour::label_id from that same run.
w	Frame width in pixels.
h	Frame height in pixels.

Returns

New image, transparent everywhere outside the selected label's silhouette.

Definition at line 450 of file VisionComponentsDispatch.cpp.

{
    auto& component_contours = contexts.component_contours;
    auto& foundry = Portal::Graphics::get_shader_foundry();
 
    component_contours.swap_shader({
        .shader_path = "vision_label_select.comp.spv",
        .workgroup_size = k_wg2d,
        .push_constant_size = sizeof(SelectLabelPC),
    });
    component_contours.stage_image_at(1, source, GpuBufferBinding::ElementType::IMAGE_SAMPLED);
    component_contours.set_push_constants(SelectLabelPC {
        .target_label = target_label, .width = w, .height = h });
    component_contours.prepare_output_image(w, h);
    component_contours.set_output_dimensions(w, h);
 
    const auto fence = component_contours.dispatch_async({});
    component_contours.clear_output_dimensions();
    foundry.wait_for_fence(fence);
    foundry.release_fence(fence);
 
    return component_contours.get_output_image(0);
}

References MayaFlux::Yantra::VisionGpuContexts::component_contours, fence, MayaFlux::Portal::Graphics::get_shader_foundry(), h, MayaFlux::Yantra::VisionInternal::k_wg2d, and MayaFlux::Yantra::VisionInternal::SelectLabelPC::target_label.

Referenced by MayaFlux::Yantra::VisionExtractor::mask().

Here is the call graph for this function:

Here is the caller graph for this function: