NVIDIA DeepStream SDK API Reference
9.2 Release
Functions
sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp File Reference

Go to the source code of this file.

Functions

static void createOnnxProfile4NanoSamDecoder (nvinfer1::INetworkDefinition *network, nvinfer1::IBuilder *builder, nvinfer1::IBuilderConfig *config, int batchSize)
 Configure dynamic input shapes for the NanoSAM decoder. More...
 
static void setRuntimeDim4NanoSamDecoder (TrtModule *trtModule, const std::vector< int > &sizes)
 Set runtime input shapes for the NanoSAM decoder. More...
 
static void createOnnxProfile4EVSamEncoder (nvinfer1::INetworkDefinition *network, nvinfer1::IBuilder *builder, nvinfer1::IBuilderConfig *config, int batchSize)
 Configure dynamic input shapes for the EfficientViT-SAM encoder. More...
 
static void setRuntimeDim4EVSamEncoder (TrtModule *trtModule, const std::vector< int > &sizes)
 Set runtime input shapes for the EfficientViT-SAM encoder. More...
 
static void createOnnxProfile4EVSamDecoder (nvinfer1::INetworkDefinition *network, nvinfer1::IBuilder *builder, nvinfer1::IBuilderConfig *config, int batchSize)
 Configure dynamic input shapes for the EfficientViT-SAM decoder. More...
 
static void setRuntimeDim4EVSamDecoder (TrtModule *trtModule, const std::vector< int > &sizes)
 Set runtime input shapes for the EfficientViT-SAM decoder. More...
 
static void createOnnxProfile4Sam2MaskDecoder (nvinfer1::INetworkDefinition *network, nvinfer1::IBuilder *builder, nvinfer1::IBuilderConfig *config, int batchSize)
 Configure dynamic input shapes for the SAM2 mask decoder. More...
 
static void setRuntimeDim4Sam2MaskDecoder (TrtModule *trtModule, const std::vector< int > &sizes)
 Set runtime input shapes for the SAM2 mask decoder. More...
 
static void createOnnxProfile4Sam2MemAttn (nvinfer1::INetworkDefinition *network, nvinfer1::IBuilder *builder, nvinfer1::IBuilderConfig *config, int batchSize)
 Configure dynamic input shapes for SAM2 memory attention. More...
 
static void setRuntimeDim4Sam2MemAttn (TrtModule *trtModule, const std::vector< int > &sizes)
 Set runtime input shapes for SAM2 memory attention. More...
 
static void createOnnxProfile4Sam2MemEncoder (nvinfer1::INetworkDefinition *network, nvinfer1::IBuilder *builder, nvinfer1::IBuilderConfig *config, int batchSize)
 Configure dynamic input shapes for the SAM2 memory encoder. More...
 
static void setRuntimeDim4Sam2MemEncoder (TrtModule *trtModule, const std::vector< int > &sizes)
 Set runtime input shapes for the SAM2 memory encoder. More...
 

Function Documentation

◆ createOnnxProfile4EVSamDecoder()

static void createOnnxProfile4EVSamDecoder ( nvinfer1::INetworkDefinition *  network,
nvinfer1::IBuilder *  builder,
nvinfer1::IBuilderConfig *  config,
int  batchSize 
)
static

Configure dynamic input shapes for the EfficientViT-SAM decoder.

Parameters
[in,out]networkTensorRT network definition.
[in]builderTensorRT engine builder.
[in,out]configTensorRT builder configuration.
[in]batchSizeMaximum runtime batch size.

Definition at line 113 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

◆ createOnnxProfile4EVSamEncoder()

static void createOnnxProfile4EVSamEncoder ( nvinfer1::INetworkDefinition *  network,
nvinfer1::IBuilder *  builder,
nvinfer1::IBuilderConfig *  config,
int  batchSize 
)
static

Configure dynamic input shapes for the EfficientViT-SAM encoder.

Parameters
[in,out]networkTensorRT network definition.
[in]builderTensorRT engine builder.
[in,out]configTensorRT builder configuration.
[in]batchSizeMaximum runtime batch size.

Definition at line 67 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

◆ createOnnxProfile4NanoSamDecoder()

static void createOnnxProfile4NanoSamDecoder ( nvinfer1::INetworkDefinition *  network,
nvinfer1::IBuilder *  builder,
nvinfer1::IBuilderConfig *  config,
int  batchSize 
)
static

Configure dynamic input shapes for the NanoSAM decoder.

Parameters
[in,out]networkTensorRT network definition.
[in]builderTensorRT engine builder.
[in,out]configTensorRT builder configuration.
[in]batchSizeMaximum runtime batch size.

Definition at line 27 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

◆ createOnnxProfile4Sam2MaskDecoder()

static void createOnnxProfile4Sam2MaskDecoder ( nvinfer1::INetworkDefinition *  network,
nvinfer1::IBuilder *  builder,
nvinfer1::IBuilderConfig *  config,
int  batchSize 
)
static

Configure dynamic input shapes for the SAM2 mask decoder.

Parameters
[in,out]networkTensorRT network definition.
[in]builderTensorRT engine builder.
[in,out]configTensorRT builder configuration.
[in]batchSizeMaximum runtime batch size.

Definition at line 153 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

◆ createOnnxProfile4Sam2MemAttn()

static void createOnnxProfile4Sam2MemAttn ( nvinfer1::INetworkDefinition *  network,
nvinfer1::IBuilder *  builder,
nvinfer1::IBuilderConfig *  config,
int  batchSize 
)
static

Configure dynamic input shapes for SAM2 memory attention.

Parameters
[in,out]networkTensorRT network definition.
[in]builderTensorRT engine builder.
[in,out]configTensorRT builder configuration.
[in]batchSizeMaximum runtime batch size.

m_memory_obj_ptr_d [b x 1 x 256] to [b x 16 x 256].

m_memory_maskmem_features_d [b x 1 x 64 x 64 x 64] to [b x 7 x 64 x 64 x 64].

Positional encoding shape: [batch, mask memories * 4096 + object pointers * 4, 64]. This gives sequence lengths of 4100 at minimum and 28736 at maximum.

Definition at line 212 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

◆ createOnnxProfile4Sam2MemEncoder()

static void createOnnxProfile4Sam2MemEncoder ( nvinfer1::INetworkDefinition *  network,
nvinfer1::IBuilder *  builder,
nvinfer1::IBuilderConfig *  config,
int  batchSize 
)
static

Configure dynamic input shapes for the SAM2 memory encoder.

Parameters
[in,out]networkTensorRT network definition.
[in]builderTensorRT engine builder.
[in,out]configTensorRT builder configuration.
[in]batchSizeMaximum runtime batch size.

Definition at line 276 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

◆ setRuntimeDim4EVSamDecoder()

static void setRuntimeDim4EVSamDecoder ( TrtModule *  trtModule,
const std::vector< int > &  sizes 
)
static

Set runtime input shapes for the EfficientViT-SAM decoder.

Parameters
[in,out]trtModuleTensorRT module whose runtime dimensions are configured.
[in]sizesRuntime tensor dimensions.

Definition at line 139 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

References TrtModule::m_context, TrtModule::m_cuStream, and TrtModule::m_engine.

◆ setRuntimeDim4EVSamEncoder()

static void setRuntimeDim4EVSamEncoder ( TrtModule *  trtModule,
const std::vector< int > &  sizes 
)
static

Set runtime input shapes for the EfficientViT-SAM encoder.

Parameters
[in,out]trtModuleTensorRT module whose runtime dimensions are configured.
[in]sizesRuntime tensor dimensions.

Definition at line 91 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

References TrtModule::m_context, and TrtModule::m_engine.

◆ setRuntimeDim4NanoSamDecoder()

static void setRuntimeDim4NanoSamDecoder ( TrtModule *  trtModule,
const std::vector< int > &  sizes 
)
static

Set runtime input shapes for the NanoSAM decoder.

Parameters
[in,out]trtModuleTensorRT module whose runtime dimensions are configured.
[in]sizesRuntime tensor dimensions.

Definition at line 53 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

References TrtModule::m_context, TrtModule::m_cuStream, and TrtModule::m_engine.

◆ setRuntimeDim4Sam2MaskDecoder()

static void setRuntimeDim4Sam2MaskDecoder ( TrtModule *  trtModule,
const std::vector< int > &  sizes 
)
static

Set runtime input shapes for the SAM2 mask decoder.

Parameters
[in,out]trtModuleTensorRT module whose runtime dimensions are configured.
[in]sizesRuntime tensor dimensions.

Definition at line 191 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

References TrtModule::m_context, TrtModule::m_cuStream, and TrtModule::m_engine.

◆ setRuntimeDim4Sam2MemAttn()

static void setRuntimeDim4Sam2MemAttn ( TrtModule *  trtModule,
const std::vector< int > &  sizes 
)
static

Set runtime input shapes for SAM2 memory attention.

Parameters
[in,out]trtModuleTensorRT module whose runtime dimensions are configured.
[in]sizesRuntime tensor dimensions.

Definition at line 255 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

References TrtModule::m_context, TrtModule::m_cuStream, and TrtModule::m_engine.

◆ setRuntimeDim4Sam2MemEncoder()

static void setRuntimeDim4Sam2MemEncoder ( TrtModule *  trtModule,
const std::vector< int > &  sizes 
)
static

Set runtime input shapes for the SAM2 memory encoder.

Parameters
[in,out]trtModuleTensorRT module whose runtime dimensions are configured.
[in]sizesRuntime tensor dimensions.

Definition at line 305 of file sources/libs/nvmultiobjecttracker/include/modules/Segmenter/SegmenterProfile.hpp.

References TrtModule::m_context, TrtModule::m_cuStream, and TrtModule::m_engine.