52 lines
1.5 KiB
C++
52 lines
1.5 KiB
C++
//
|
|
// Created by DefTruth on 2021/3/14.
|
|
//
|
|
|
|
#include "pfld.h"
|
|
#include "lite/ort/core/ort_utils.h"
|
|
|
|
using ortcv::PFLD;
|
|
|
|
Ort::Value PFLD::transform(const cv::Mat &mat)
|
|
{
|
|
cv::Mat canvas;
|
|
cv::resize(mat, canvas, cv::Size(input_node_dims.at(3),
|
|
input_node_dims.at(2)));
|
|
// (1,3,112,112)
|
|
ortcv::utils::transform::normalize_inplace(canvas, mean_val, scale_val); // float32
|
|
|
|
return ortcv::utils::transform::create_tensor(
|
|
canvas, input_node_dims, memory_info_handler,
|
|
input_values_handler, ortcv::utils::transform::CHW);
|
|
}
|
|
|
|
void PFLD::detect(const cv::Mat &mat, types::Landmarks &landmarks)
|
|
{
|
|
if (mat.empty()) return;
|
|
float img_height = static_cast<float>(mat.rows);
|
|
float img_width = static_cast<float>(mat.cols);
|
|
|
|
// 1. make input tensor
|
|
Ort::Value input_tensor = this->transform(mat);
|
|
// 2. inference
|
|
auto output_tensors = ort_session->Run(
|
|
Ort::RunOptions{nullptr}, input_node_names.data(),
|
|
&input_tensor, 1, output_node_names.data(), num_outputs
|
|
);
|
|
// 3. fetch landmarks.
|
|
Ort::Value &landmarks_norm = output_tensors.at(1); // (1,106*2)
|
|
auto landmark_dims = output_node_dims.at(1);
|
|
const unsigned int num_landmarks = landmark_dims.at(1);
|
|
|
|
for (unsigned int i = 0; i < num_landmarks; i += 2)
|
|
{
|
|
float x = landmarks_norm.At<float>({0, i});
|
|
float y = landmarks_norm.At<float>({0, i + 1});
|
|
|
|
x = std::min(std::max(0.f, x), 1.0f);
|
|
y = std::min(std::max(0.f, y), 1.0f);
|
|
|
|
landmarks.points.push_back(cv::Point2f(x * img_width, y * img_height));
|
|
}
|
|
landmarks.flag = true;
|
|
} |