object_detector.h 4.3 KB
Newer Older
Q
qingqing01 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
//   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,
C
cnn 已提交
53
                     const std::vector<std::string>& lables,
C
cnn 已提交
54 55
                     const std::vector<int>& colormap,
                     const bool is_rbox);
Q
qingqing01 已提交
56 57 58 59 60 61


class ObjectDetector {
 public:
  explicit ObjectDetector(const std::string& model_dir, 
                          bool use_gpu=false,
G
Guanghua Yu 已提交
62 63
                          bool use_mkldnn=false,
                          int cpu_threads=1,
Q
qingqing01 已提交
64
                          const std::string& run_mode="fluid",
65
                          const int batch_size=1,
66 67 68 69
                          const int gpu_id=0,
                          bool use_dynamic_shape=false,
                          const int trt_min_shape=1,
                          const int trt_max_shape=1280,
G
Guanghua Yu 已提交
70 71 72 73 74 75 76 77 78 79 80 81
                          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;
Q
qingqing01 已提交
82
    config_.load_config(model_dir);
G
Guanghua Yu 已提交
83
    this->min_subgraph_size_ = config_.min_subgraph_size_;
Q
qingqing01 已提交
84 85 86
    threshold_ = config_.draw_threshold_;
    image_shape_ = config_.image_shape_;
    preprocessor_.Init(config_.preprocess_info_, image_shape_);
87
    LoadModel(model_dir, batch_size, run_mode);
Q
qingqing01 已提交
88 89 90 91 92 93
  }

  // Load Paddle inference model
  void LoadModel(
    const std::string& model_dir,
    const int batch_size = 1,
G
Guanghua Yu 已提交
94
    const std::string& run_mode = "fluid");
Q
qingqing01 已提交
95 96

  // Run predictor
C
cnn 已提交
97
  void Predict(const std::vector<cv::Mat> imgs,
Q
qingqing01 已提交
98 99 100
      const double threshold = 0.5,
      const int warmup = 0,
      const int repeats = 1,
G
Guanghua Yu 已提交
101
      std::vector<ObjectResult>* result = nullptr,
C
cnn 已提交
102
      std::vector<int>* bbox_num = nullptr,
G
Guanghua Yu 已提交
103
      std::vector<double>* times = nullptr);
Q
qingqing01 已提交
104 105 106 107 108 109 110

  // Get Model Label list
  const std::vector<std::string>& GetLabelList() const {
    return config_.label_list_;
  }

 private:
G
Guanghua Yu 已提交
111 112 113 114 115 116 117 118 119 120
  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;
Q
qingqing01 已提交
121 122 123 124
  // Preprocess image and copy data to input buffer
  void Preprocess(const cv::Mat& image_mat);
  // Postprocess result
  void Postprocess(
C
cnn 已提交
125
      const std::vector<cv::Mat> mats,
C
cnn 已提交
126
      std::vector<ObjectResult>* result,
C
cnn 已提交
127
      std::vector<int> bbox_num,
C
cnn 已提交
128
      bool is_rbox);
Q
qingqing01 已提交
129 130 131 132 133

  std::shared_ptr<Predictor> predictor_;
  Preprocessor preprocessor_;
  ImageBlob inputs_;
  std::vector<float> output_data_;
C
cnn 已提交
134
  std::vector<int> out_bbox_num_data_;
Q
qingqing01 已提交
135 136 137 138 139 140
  float threshold_;
  ConfigPaser config_;
  std::vector<int> image_shape_;
};

}  // namespace PaddleDetection