onnxtensorinference

This element runs an ONNX model using tensors attached to the input buffer. Unlike onnxinference, it takes tensors from GstTensorMeta or GstTensorListMeta as model inputs instead of processing the input media buffer. Inference results are attached to the buffer as tensor metadata.

Input tensors can be produced by an upstream preprocessor or another inference element, allowing multiple inference models to be chained. GstTensorListMeta can also be used to process multiple tensor sets from the same buffer.

Hierarchy

GObject
    ╰──GInitiallyUnowned
        ╰──GstObject
            ╰──GstElement
                ╰──GstBaseTransform
                    ╰──GstOnnxBaseInference
                        ╰──onnxtensorinference

Factory details

Authors: – Seungha Yang

Classification: – Filter

Rank – none

Plugin – onnx

Package – GStreamer Bad Plug-ins

Pad Templates

sink

ANY

Presence – always

Direction – sink

Object type – GstPad


src

ANY

Presence – always

Direction – src

Object type – GstPad


Properties

remove-tensor-meta

“remove-tensor-meta” gboolean

Remove used input tensor meta after inference

Flags : Read / Write

Default value : true


The results of the search are