Struct TensorMeta 

pub struct TensorMeta(/* private fields */);

The inference outputs attached to a frame, for the elements that keep the picture on the wire instead of replacing it with the tensor. That is what lets inference -> post-process -> overlay stay one straight chain: the frame reaching the overlay is still the video the model saw.

Holds every tensor on the frame, since a FrameMetaSet keys by concrete type and would otherwise let a second model’s output replace the first. These are not detections; those are the post-processor’s AnalyticsMeta.

Implementations

impl TensorMeta

pub fn new() -> Self

An empty set, the starting point a producer pushes onto.

pub fn push(&mut self, tensor: NamedTensor)

Add one output, replacing any earlier tensor of the same name so a re-run of the same stage updates rather than accumulates.

pub fn iter(&self) -> impl Iterator<Item = &NamedTensor>

Every attached tensor, in the order the stages produced them.

pub fn get(&self, name: &str) -> Option<&NamedTensor>

The tensor named name.

pub fn only(&self) -> Option<&NamedTensor>

The tensor when the frame carries exactly one, so the ordinary single-model pipeline needs no names. None when there are several: the consumer must then say which it wants rather than be handed an arbitrary one.

pub fn is_empty(&self) -> bool

Trait Implementations

impl Clone for TensorMeta

fn clone(&self) -> TensorMeta

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for TensorMeta

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for TensorMeta

fn default() -> TensorMeta

Returns the “default value” for a type.

impl FrameMeta for TensorMeta

fn propagate(&self, transform: Transform) -> Propagation

The tensors describe the picture the models saw, so a re-encode ends their usefulness the way it ends an analytics graph’s. A scale or crop leaves them readable: what they mean is fixed by the model input size the post-processor normalizes against, not the frame’s current size.

fn as_any(&self) -> &dyn Any

fn as_any_mut(&mut self) -> &mut dyn Any

fn clone_box(&self) -> Box<dyn FrameMeta>

A boxed deep copy of this meta, the GstMeta copy_func analog. Backs the copy-on-write of a shared meta when a tee branch mutates it (see FrameMetaSet::get_mut); each FrameMeta impl is its own concrete type, so the duplication can only be expressed on the trait.

impl PartialEq for TensorMeta

fn eq(&self, other: &TensorMeta) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for TensorMeta