VvasInferPrediction - VvasInferPrediction - 6.3 English - XD056

VvasInferPrediction is the VVAS-level prediction structure embedded inside GstInferencePrediction. It holds the bounding box, inference result, and the tree linkage for cascade (multi-level) inference.

Definition

typedef struct {
  uint64_t prediction_id;
  bool enabled;
  VvasTreeNode *node;
  bool scaled_to_root;
  char *model_path;
  VvasList *tensors;
  TensorBuf *tb_ref;
  VvasInferPredictionPriv priv;
  VvasInferResult *infer_result;
  int width;
  int height;
} VvasInferPrediction;

Members

prediction_id

Unique identifier for this prediction node.

enabled

When true, this prediction is used for further (cascade) inference. Set to false to skip this node without removing it from the tree.

node

Address of the tree node used to build the hierarchical prediction tree. Child predictions are appended via vvas_inferprediction_append().

scaled_to_root

Indicates whether the bounding-box coordinates have been scaled to the root-node (original frame) resolution.

model_path

Path to the model that produced this prediction.

tensors

List of raw output tensors (TensorBuf) from the model. Used by the post-processing library to extract structured results.

tb_ref

Reference to the TensorBuf from which the inference result was extracted; maintains tensor lifetime while results are in use.

priv

Application-private data slot. The application provides free and copy callbacks to manage its own associated data.

infer_result

Structured inference result populated by the post-processing library. See VvasInferResult and VvasInferResultType.

width / height

Dimensions of the input image for which this prediction was generated.