NVIDIA DeepStream SDK API Reference

6.4 Release
NvDsInferTensorMeta Struct Reference

Detailed Description

Holds the raw tensor output information for one frame / one object.

The "nvinfer" plugins adds this meta when the "output-tensor-meta" property of the element instance is set to TRUE.

This meta data is added as NvDsUserMeta to the frame_user_meta_list of the corresponding frame_meta or object_user_meta_list of the corresponding object with the meta_type set to NVDSINFER_TENSOR_OUTPUT_META.

Definition at line 70 of file gstnvdsinfer.h.

Collaboration diagram for NvDsInferTensorMeta:

Data Fields

guint unique_id
 Unique ID of the gst-nvinfer instance which attached this meta. More...
 
guint num_output_layers
 Number of bound output layers. More...
 
NvDsInferLayerInfooutput_layers_info
 Pointer to the array containing information for the bound output layers. More...
 
void ** out_buf_ptrs_host
 Array of pointers to the output host buffers for the batch / frame / object. More...
 
void ** out_buf_ptrs_dev
 Array of pointers to the output device buffers for the batch / frame / object. More...
 
gint gpu_id
 GPU device ID on which the device buffers have been allocated. More...
 
void * priv_data
 Private data used for the meta producer's internal memory management. More...
 
NvDsInferNetworkInfo network_info
 Network information for the model specified for the nvinfer element instance. More...
 
gboolean maintain_aspect_ratio
 Whether aspect ratio was maintained while scaling to network resolution. More...
 

Field Documentation

◆ gpu_id

gint NvDsInferTensorMeta::gpu_id

GPU device ID on which the device buffers have been allocated.

Definition at line 85 of file gstnvdsinfer.h.

◆ maintain_aspect_ratio

gboolean NvDsInferTensorMeta::maintain_aspect_ratio

Whether aspect ratio was maintained while scaling to network resolution.

Definition at line 91 of file gstnvdsinfer.h.

◆ network_info

NvDsInferNetworkInfo NvDsInferTensorMeta::network_info

Network information for the model specified for the nvinfer element instance.

Definition at line 89 of file gstnvdsinfer.h.

◆ num_output_layers

guint NvDsInferTensorMeta::num_output_layers

Number of bound output layers.

Definition at line 75 of file gstnvdsinfer.h.

◆ out_buf_ptrs_dev

void** NvDsInferTensorMeta::out_buf_ptrs_dev

Array of pointers to the output device buffers for the batch / frame / object.

Definition at line 83 of file gstnvdsinfer.h.

◆ out_buf_ptrs_host

void** NvDsInferTensorMeta::out_buf_ptrs_host

Array of pointers to the output host buffers for the batch / frame / object.

Definition at line 81 of file gstnvdsinfer.h.

◆ output_layers_info

NvDsInferLayerInfo* NvDsInferTensorMeta::output_layers_info

Pointer to the array containing information for the bound output layers.

Size of the array will be equal to num_output_layers. Pointers inside the NvDsInferLayerInfo structure are not valid for this array.

Definition at line 79 of file gstnvdsinfer.h.

◆ priv_data

void* NvDsInferTensorMeta::priv_data

Private data used for the meta producer's internal memory management.

Definition at line 87 of file gstnvdsinfer.h.

◆ unique_id

guint NvDsInferTensorMeta::unique_id

Unique ID of the gst-nvinfer instance which attached this meta.

Definition at line 73 of file gstnvdsinfer.h.


The documentation for this struct was generated from the following file: