ocr_rec.cpp 6.3 KB
Newer Older
littletomatodonkey's avatar
littletomatodonkey 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
// 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.

#include <include/ocr_rec.h>

namespace PaddleOCR {

void CRNNRecognizer::Run(std::vector<std::vector<std::vector<int>>> boxes,
                         cv::Mat &img) {
  cv::Mat srcimg;
  img.copyTo(srcimg);
  cv::Mat crop_img;
  cv::Mat resize_img;

  std::cout << "The predicted text is :" << std::endl;
  int index = 0;
  for (int i = boxes.size() - 1; i >= 0; i--) {
littletomatodonkey's avatar
littletomatodonkey 已提交
29
    crop_img = GetRotateCropImage(srcimg, boxes[i]);
littletomatodonkey's avatar
littletomatodonkey 已提交
30 31 32 33 34 35 36 37

    float wh_ratio = float(crop_img.cols) / float(crop_img.rows);

    this->resize_op_.Run(crop_img, resize_img, wh_ratio);

    this->normalize_op_.Run(&resize_img, this->mean_, this->scale_,
                            this->is_scale_);

littletomatodonkey's avatar
littletomatodonkey 已提交
38
    std::vector<float> input(1 * 3 * resize_img.rows * resize_img.cols, 0.0f);
littletomatodonkey's avatar
littletomatodonkey 已提交
39

littletomatodonkey's avatar
littletomatodonkey 已提交
40
    this->permute_op_.Run(&resize_img, input.data());
littletomatodonkey's avatar
littletomatodonkey 已提交
41 42 43 44

    auto input_names = this->predictor_->GetInputNames();
    auto input_t = this->predictor_->GetInputTensor(input_names[0]);
    input_t->Reshape({1, 3, resize_img.rows, resize_img.cols});
littletomatodonkey's avatar
littletomatodonkey 已提交
45
    input_t->copy_from_cpu(input.data());
littletomatodonkey's avatar
littletomatodonkey 已提交
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91

    this->predictor_->ZeroCopyRun();

    std::vector<int64_t> rec_idx;
    auto output_names = this->predictor_->GetOutputNames();
    auto output_t = this->predictor_->GetOutputTensor(output_names[0]);
    auto rec_idx_lod = output_t->lod();
    auto shape_out = output_t->shape();
    int out_num = std::accumulate(shape_out.begin(), shape_out.end(), 1,
                                  std::multiplies<int>());

    rec_idx.resize(out_num);
    output_t->copy_to_cpu(rec_idx.data());

    std::vector<int> pred_idx;
    for (int n = int(rec_idx_lod[0][0]); n < int(rec_idx_lod[0][1]); n++) {
      pred_idx.push_back(int(rec_idx[n]));
    }

    if (pred_idx.size() < 1e-3)
      continue;

    index += 1;
    std::cout << index << "\t";
    for (int n = 0; n < pred_idx.size(); n++) {
      std::cout << label_list_[pred_idx[n]];
    }

    std::vector<float> predict_batch;
    auto output_t_1 = this->predictor_->GetOutputTensor(output_names[1]);

    auto predict_lod = output_t_1->lod();
    auto predict_shape = output_t_1->shape();
    int out_num_1 = std::accumulate(predict_shape.begin(), predict_shape.end(),
                                    1, std::multiplies<int>());

    predict_batch.resize(out_num_1);
    output_t_1->copy_to_cpu(predict_batch.data());

    int argmax_idx;
    int blank = predict_shape[1];
    float score = 0.f;
    int count = 0;
    float max_value = 0.0f;

    for (int n = predict_lod[0][0]; n < predict_lod[0][1] - 1; n++) {
littletomatodonkey's avatar
littletomatodonkey 已提交
92 93
      argmax_idx =
          int(Utility::argmax(&predict_batch[n * predict_shape[1]],
littletomatodonkey's avatar
littletomatodonkey 已提交
94 95 96 97 98 99 100 101 102 103 104 105 106 107
                              &predict_batch[(n + 1) * predict_shape[1]]));
      max_value =
          float(*std::max_element(&predict_batch[n * predict_shape[1]],
                                  &predict_batch[(n + 1) * predict_shape[1]]));
      if (blank - 1 - argmax_idx > 1e-5) {
        score += max_value;
        count += 1;
      }
    }
    score /= count;
    std::cout << "\tscore: " << score << std::endl;
  }
}

littletomatodonkey's avatar
littletomatodonkey 已提交
108
void CRNNRecognizer::LoadModel(const std::string &model_dir) {
littletomatodonkey's avatar
littletomatodonkey 已提交
109 110 111
  AnalysisConfig config;
  config.SetModel(model_dir + "/model", model_dir + "/params");

littletomatodonkey's avatar
littletomatodonkey 已提交
112 113 114 115
  if (this->use_gpu_) {
    config.EnableUseGpu(this->gpu_mem_, this->gpu_id_);
  } else {
    config.DisableGpu();
littletomatodonkey's avatar
littletomatodonkey 已提交
116 117 118
    if (this->use_mkldnn_) {
      config.EnableMKLDNN();
    }
littletomatodonkey's avatar
littletomatodonkey 已提交
119 120
    config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_);
  }
littletomatodonkey's avatar
littletomatodonkey 已提交
121

littletomatodonkey's avatar
littletomatodonkey 已提交
122
  // false for zero copy tensor
littletomatodonkey's avatar
littletomatodonkey 已提交
123
  config.SwitchUseFeedFetchOps(false);
littletomatodonkey's avatar
littletomatodonkey 已提交
124
  // true for multiple input
littletomatodonkey's avatar
littletomatodonkey 已提交
125
  config.SwitchSpecifyInputNames(true);
littletomatodonkey's avatar
littletomatodonkey 已提交
126 127 128 129

  config.SwitchIrOptim(true);

  config.EnableMemoryOptim();
littletomatodonkey's avatar
littletomatodonkey 已提交
130 131 132 133

  this->predictor_ = CreatePaddlePredictor(config);
}

littletomatodonkey's avatar
littletomatodonkey 已提交
134 135
cv::Mat CRNNRecognizer::GetRotateCropImage(const cv::Mat &srcimage,
                                           std::vector<std::vector<int>> box) {
littletomatodonkey's avatar
littletomatodonkey 已提交
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189
  cv::Mat image;
  srcimage.copyTo(image);
  std::vector<std::vector<int>> points = box;

  int x_collect[4] = {box[0][0], box[1][0], box[2][0], box[3][0]};
  int y_collect[4] = {box[0][1], box[1][1], box[2][1], box[3][1]};
  int left = int(*std::min_element(x_collect, x_collect + 4));
  int right = int(*std::max_element(x_collect, x_collect + 4));
  int top = int(*std::min_element(y_collect, y_collect + 4));
  int bottom = int(*std::max_element(y_collect, y_collect + 4));

  cv::Mat img_crop;
  image(cv::Rect(left, top, right - left, bottom - top)).copyTo(img_crop);

  for (int i = 0; i < points.size(); i++) {
    points[i][0] -= left;
    points[i][1] -= top;
  }

  int img_crop_width = int(sqrt(pow(points[0][0] - points[1][0], 2) +
                                pow(points[0][1] - points[1][1], 2)));
  int img_crop_height = int(sqrt(pow(points[0][0] - points[3][0], 2) +
                                 pow(points[0][1] - points[3][1], 2)));

  cv::Point2f pts_std[4];
  pts_std[0] = cv::Point2f(0., 0.);
  pts_std[1] = cv::Point2f(img_crop_width, 0.);
  pts_std[2] = cv::Point2f(img_crop_width, img_crop_height);
  pts_std[3] = cv::Point2f(0.f, img_crop_height);

  cv::Point2f pointsf[4];
  pointsf[0] = cv::Point2f(points[0][0], points[0][1]);
  pointsf[1] = cv::Point2f(points[1][0], points[1][1]);
  pointsf[2] = cv::Point2f(points[2][0], points[2][1]);
  pointsf[3] = cv::Point2f(points[3][0], points[3][1]);

  cv::Mat M = cv::getPerspectiveTransform(pointsf, pts_std);

  cv::Mat dst_img;
  cv::warpPerspective(img_crop, dst_img, M,
                      cv::Size(img_crop_width, img_crop_height),
                      cv::BORDER_REPLICATE);

  if (float(dst_img.rows) >= float(dst_img.cols) * 1.5) {
    cv::Mat srcCopy = cv::Mat(dst_img.rows, dst_img.cols, dst_img.depth());
    cv::transpose(dst_img, srcCopy);
    cv::flip(srcCopy, srcCopy, 0);
    return srcCopy;
  } else {
    return dst_img;
  }
}

} // namespace PaddleOCR