vitis::ai::RetinaFacePostProcess

Class of the retinaface post-process. It initializes the parameters once instead of computing them each time the program executes.

Quick Function Reference

The following table lists all the functions defined in the vitis::ai::RetinaFacePostProcess class:

Table 1. Quick Function Reference
Type	Name	Arguments
std::unique_ptr< `RetinaFacePostProcess` >	create	const std::vector< vitis::ai::library::InputTensor > & input_tensors const std::vector< vitis::ai::library::OutputTensor > & output_tensors const vitis::ai::proto::DpuModelParam & config
std::vector< `RetinaFaceResult` >	retinaface_post_process	void

create

Create a RetinaFacePostProcess object.

Prototype

std::unique_ptr< RetinaFacePostProcess > create(const std::vector< vitis::ai::library::InputTensor > &input_tensors, const std::vector< vitis::ai::library::OutputTensor > &output_tensors, const vitis::ai::proto::DpuModelParam &config);

Parameters

The following table lists the create function arguments.

Table 2. create Arguments
Type	Name	Description
const std::vector< vitis::ai::library::InputTensor > &	input_tensors	A vector of all input-tensors in the network. Usage: input_tensors[input_tensor_index].
const std::vector< vitis::ai::library::OutputTensor > &	output_tensors	A vector of all output-tensors in the network. Usage: output_tensors[output_index].
const vitis::ai::proto::DpuModelParam &	config	The DPU model configuration information.

Returns

A unique pointer of RetinaFacePostProcess.

retinaface_post_process

The batch mode post-processing function of the retinaface network.

Prototype

std::vector< RetinaFaceResult > retinaface_post_process(size_t batch_size)=0;

Returns

The vector of struct of RetinaFaceResult.