NVIDIA DeepStream SDK API Reference

9.2 Release
9.2/sources/gst-plugins/gst-nvinfereval/gstnvinfereval.h File Reference

Go to the source code of this file.

Data Structures

struct  _GstNvInferEval
 GstNvInferEval: More...
 
struct  _GstNvInferEvalClass
 

Macros

#define GST_TYPE_NVINFEREVAL   (gst_nv_infer_eval_get_type ())
 SECTION:element-nvinfereval. More...
 
#define GST_NVINFEREVAL(obj)   (G_TYPE_CHECK_INSTANCE_CAST ((obj), GST_TYPE_NVINFEREVAL, GstNvInferEval))
 
#define GST_NVINFEREVAL_CLASS(klass)   (G_TYPE_CHECK_CLASS_CAST ((klass), GST_TYPE_NVINFEREVAL, GstNvInferEvalClass))
 
#define GST_IS_NVINFEREVAL(obj)   (G_TYPE_CHECK_INSTANCE_TYPE ((obj), GST_TYPE_NVINFEREVAL))
 
#define GST_IS_NVINFEREVAL_CLASS(klass)   (G_TYPE_CHECK_CLASS_TYPE ((klass), GST_TYPE_NVINFEREVAL))
 

Typedefs

typedef struct _GstNvInferEval GstNvInferEval
 
typedef struct _GstNvInferEvalClass GstNvInferEvalClass
 

Functions

GType gst_nv_infer_eval_get_type (void)
 

Macro Definition Documentation

◆ GST_IS_NVINFEREVAL

#define GST_IS_NVINFEREVAL (   obj)    (G_TYPE_CHECK_INSTANCE_TYPE ((obj), GST_TYPE_NVINFEREVAL))

◆ GST_IS_NVINFEREVAL_CLASS

#define GST_IS_NVINFEREVAL_CLASS (   klass)    (G_TYPE_CHECK_CLASS_TYPE ((klass), GST_TYPE_NVINFEREVAL))

◆ GST_NVINFEREVAL

#define GST_NVINFEREVAL (   obj)    (G_TYPE_CHECK_INSTANCE_CAST ((obj), GST_TYPE_NVINFEREVAL, GstNvInferEval))

◆ GST_NVINFEREVAL_CLASS

#define GST_NVINFEREVAL_CLASS (   klass)    (G_TYPE_CHECK_CLASS_CAST ((klass), GST_TYPE_NVINFEREVAL, GstNvInferEvalClass))

◆ GST_TYPE_NVINFEREVAL

#define GST_TYPE_NVINFEREVAL   (gst_nv_infer_eval_get_type ())

SECTION:element-nvinfereval.

nvinfereval

A pass-through DeepStream GStreamer element that evaluates model accuracy live against ground-truth files. Buffers flow downstream UNCHANGED.

Two operating modes, auto-detected at READY->PAUSED by walking upstream through transparent elements (queue, queue2, identity, capsfilter):

Mode A — Single model nvinfer -> [queue?] -> nvinfereval -> sink No InferenceProvenanceMeta on frames. Evaluates one model vs GT. No promotion logic.

Mode B — Multi-model A/B evaluation + promotion nvmodelmux -> [queue?] -> nvinfereval -> sink InferenceProvenanceMeta present per frame (role / model_name / model_version). Any number of model_names may coexist — each gets its own isolated epoch counter, GT index, and accumulator. Promotion is triggered independently per model_name and targets only the streams carrying that pair.

Configuration

Individual GObject properties set baseline values. If config-file is given, its sections overlay those values (config file wins for keys it defines). See configs/config_nvinfereval.txt for the full format.

Definition at line 58 of file 9.2/sources/gst-plugins/gst-nvinfereval/gstnvinfereval.h.

Typedef Documentation

◆ GstNvInferEval

◆ GstNvInferEvalClass

Function Documentation

◆ gst_nv_infer_eval_get_type()

GType gst_nv_infer_eval_get_type ( void  )