NvDsActionRecognitionExt#

Components for using the NVIDIA TAO Toolkit Action Recognition models

  • UUID: b7259b47-6df1-4774-b55f-ccd71e692c44

  • Version: 1.6.0

  • Author: NVIDIA

  • License: Proprietary

Components#

nvidia::deepstream::NvDsActionRecognition2D#

Configuration for NVIDIA TAO 2D Action Recognition model.Recognizes actions Push, Fall, Walk, Run, Ride Bike.

For more information: https://ngc.nvidia.com/catalog/models/nvidia:tao:actionrecognitionnet

Model files must be downloaded and extracted to /opt/nvidia/deepstream/deepstream/samples/models/tao_pretrained_models/actionRecognition

  • Component ID: deca38c1-5fcd-40f6-9c2a-95f4d83524ca

  • Base Type: nvidia::deepstream::INvDsInferModelConfigComponent

Parameters#

model-engine-file

Path to the model engine file. Absolute or relative to the extension directoy.

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_FILE


enable-dla

Enable DLA for inferencing

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_BOOL

  • Default: False


use-dla-core

DLA Core to use for inferencing

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_UINT64

  • Default: 0


tensor-meta-pool-size

Size of the output tensor meta pool

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_UINT64

  • Default: 6


network-mode

Data format to be used by inference.

Valid values:

0: FP32

1: INT8

2: FP16

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_UINT64

  • Default: 2


classifier-threshold

Minimum threshold label probability. The classifier outputs the label having the highest probability if it is greater than this threshold

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_FLOAT64

  • Default: 0.0

nvidia::deepstream::NvDsActionRecognition3D#

Configuration for NVIDIA TAO 3D Action Recognition model.Recognizes actions Push, Fall, Walk, Run, Ride Bike.

For more information: https://ngc.nvidia.com/catalog/models/nvidia:tao:actionrecognitionnet

Model files must be downloaded and extracted to /opt/nvidia/deepstream/deepstream/samples/models/tao_pretrained_models/actionRecognition

  • Component ID: 0ed5a99d-5445-4eeb-823b-3f6a2f77305e

  • Base Type: nvidia::deepstream::INvDsInferModelConfigComponent

Parameters#

model-engine-file

Path to the model engine file. Absolute or relative to the extension directoy.

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_FILE


enable-dla

Enable DLA for inferencing

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_BOOL

  • Default: False


use-dla-core

DLA Core to use for inferencing

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_UINT64

  • Default: 0


tensor-meta-pool-size

Size of the output tensor meta pool

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_UINT64

  • Default: 6


network-mode

Data format to be used by inference.

Valid values:

0: FP32

1: INT8

2: FP16

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_UINT64

  • Default: 2


classifier-threshold

Minimum threshold label probability. The classifier outputs the label having the highest probability if it is greater than this threshold

  • Flags: GXF_PARAMETER_FLAGS_OPTIONAL

  • Type: GXF_PARAMETER_TYPE_FLOAT64

  • Default: 0.0