PulseFocusPlatform/deploy/cpp/include/object_detector.h

139 lines
4.2 KiB
C++

// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <string>
#include <vector>
#include <memory>
#include <utility>
#include <ctime>
#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/highgui/highgui.hpp>
#include "paddle_inference_api.h" // NOLINT
#include "include/preprocess_op.h"
#include "include/config_parser.h"
using namespace paddle_infer;
namespace PaddleDetection {
// Object Detection Result
struct ObjectResult {
// Rectangle coordinates of detected object: left, right, top, down
std::vector<int> rect;
// Class id of detected object
int class_id;
// Confidence of detected object
float confidence;
};
// Generate visualization colormap for each class
std::vector<int> GenerateColorMap(int num_class);
// Visualiztion Detection Result
cv::Mat VisualizeResult(const cv::Mat& img,
const std::vector<ObjectResult>& results,
const std::vector<std::string>& lables,
const std::vector<int>& colormap,
const bool is_rbox);
class ObjectDetector {
public:
explicit ObjectDetector(const std::string& model_dir,
bool use_gpu=false,
bool use_mkldnn=false,
int cpu_threads=1,
const std::string& run_mode="fluid",
const int batch_size=1,
const int gpu_id=0,
bool use_dynamic_shape=false,
const int trt_min_shape=1,
const int trt_max_shape=1280,
const int trt_opt_shape=640,
bool trt_calib_mode=false) {
this->use_gpu_ = use_gpu;
this->gpu_id_ = gpu_id;
this->cpu_math_library_num_threads_ = cpu_threads;
this->use_mkldnn_ = use_mkldnn;
this->use_dynamic_shape_ = use_dynamic_shape;
this->trt_min_shape_ = trt_min_shape;
this->trt_max_shape_ = trt_max_shape;
this->trt_opt_shape_ = trt_opt_shape;
this->trt_calib_mode_ = trt_calib_mode;
config_.load_config(model_dir);
this->min_subgraph_size_ = config_.min_subgraph_size_;
threshold_ = config_.draw_threshold_;
preprocessor_.Init(config_.preprocess_info_);
LoadModel(model_dir, batch_size, run_mode);
}
// Load Paddle inference model
void LoadModel(
const std::string& model_dir,
const int batch_size = 1,
const std::string& run_mode = "fluid");
// Run predictor
void Predict(const std::vector<cv::Mat> imgs,
const double threshold = 0.5,
const int warmup = 0,
const int repeats = 1,
std::vector<ObjectResult>* result = nullptr,
std::vector<int>* bbox_num = nullptr,
std::vector<double>* times = nullptr);
// Get Model Label list
const std::vector<std::string>& GetLabelList() const {
return config_.label_list_;
}
private:
bool use_gpu_ = false;
int gpu_id_ = 0;
int cpu_math_library_num_threads_ = 1;
bool use_mkldnn_ = false;
int min_subgraph_size_ = 3;
bool use_dynamic_shape_ = false;
int trt_min_shape_ = 1;
int trt_max_shape_ = 1280;
int trt_opt_shape_ = 640;
bool trt_calib_mode_ = false;
// Preprocess image and copy data to input buffer
void Preprocess(const cv::Mat& image_mat);
// Postprocess result
void Postprocess(
const std::vector<cv::Mat> mats,
std::vector<ObjectResult>* result,
std::vector<int> bbox_num,
bool is_rbox);
std::shared_ptr<Predictor> predictor_;
Preprocessor preprocessor_;
ImageBlob inputs_;
std::vector<float> output_data_;
std::vector<int> out_bbox_num_data_;
float threshold_;
ConfigPaser config_;
};
} // namespace PaddleDetection