generate_reconstructed_data.py 7.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203
  1. #!/usr/bin/env python3
  2. # -*- coding: utf-8 -*-
  3. """
  4. Created on Wed Jun 19 11:47:42 2019
  5. @author: jbuisine
  6. """
  7. # main imports
  8. import sys, os, argparse
  9. import numpy as np
  10. # images processing imports
  11. from PIL import Image
  12. from ipfml.processing.segmentation import divide_in_blocks
  13. # modules imports
  14. sys.path.insert(0, '') # trick to enable import of main folder module
  15. import custom_config as cfg
  16. from modules.utils.data import get_scene_image_quality
  17. from modules.classes.Transformation import Transformation
  18. # getting configuration information
  19. config_filename = cfg.config_filename
  20. zone_folder = cfg.zone_folder
  21. min_max_filename = cfg.min_max_filename_extension
  22. # define all scenes values
  23. scenes_list = cfg.scenes_names
  24. scenes_indexes = cfg.scenes_indices
  25. path = cfg.dataset_path
  26. zones = cfg.zones_indices
  27. seuil_expe_filename = cfg.seuil_expe_filename
  28. features_choices = cfg.features_choices_labels
  29. output_data_folder = cfg.output_data_folder
  30. generic_output_file_svd = '_random.csv'
  31. def generate_data(transformation):
  32. """
  33. @brief Method which generates all .csv files from scenes
  34. @return nothing
  35. """
  36. scenes = os.listdir(path)
  37. # remove min max file from scenes folder
  38. scenes = [s for s in scenes if min_max_filename not in s]
  39. # go ahead each scenes
  40. for id_scene, folder_scene in enumerate(scenes):
  41. print(folder_scene)
  42. scene_path = os.path.join(path, folder_scene)
  43. config_file_path = os.path.join(scene_path, config_filename)
  44. with open(config_file_path, "r") as config_file:
  45. last_image_name = config_file.readline().strip()
  46. prefix_image_name = config_file.readline().strip()
  47. start_index_image = config_file.readline().strip()
  48. end_index_image = config_file.readline().strip()
  49. step_counter = int(config_file.readline().strip())
  50. # construct each zones folder name
  51. zones_folder = []
  52. features_folder = []
  53. zones_threshold = []
  54. # get zones list info
  55. for index in zones:
  56. index_str = str(index)
  57. if len(index_str) < 2:
  58. index_str = "0" + index_str
  59. current_zone = "zone"+index_str
  60. zones_folder.append(current_zone)
  61. zone_path = os.path.join(scene_path, current_zone)
  62. with open(os.path.join(zone_path, cfg.seuil_expe_filename)) as f:
  63. zones_threshold.append(int(f.readline()))
  64. # custom path for feature
  65. feature_path = os.path.join(zone_path, transformation.getName())
  66. if not os.path.exists(feature_path):
  67. os.makedirs(feature_path)
  68. # custom path for interval of reconstruction and feature
  69. feature_interval_path = os.path.join(zone_path, transformation.getTransformationPath())
  70. features_folder.append(feature_interval_path)
  71. if not os.path.exists(feature_interval_path):
  72. os.makedirs(feature_interval_path)
  73. # create for each zone the labels folder
  74. labels = [cfg.not_noisy_folder, cfg.noisy_folder]
  75. for label in labels:
  76. label_folder = os.path.join(feature_interval_path, label)
  77. if not os.path.exists(label_folder):
  78. os.makedirs(label_folder)
  79. # get all images of folder
  80. scene_images = sorted([os.path.join(scene_path, img) for img in os.listdir(scene_path) if cfg.scene_image_extension in img])
  81. number_scene_image = len(scene_images)
  82. # for each images
  83. for id_img, img_path in enumerate(scene_images):
  84. current_img = Image.open(img_path)
  85. img_blocks = divide_in_blocks(current_img, cfg.keras_img_size)
  86. current_quality_index = int(get_scene_image_quality(img_path))
  87. for id_block, block in enumerate(img_blocks):
  88. ##########################
  89. # Image computation part #
  90. ##########################
  91. # pass block to grey level
  92. output_block = transformation.getTransformedImage(block)
  93. output_block = np.array(output_block, 'uint8')
  94. # current output image
  95. output_block_img = Image.fromarray(output_block)
  96. label_path = features_folder[id_block]
  97. # get label folder for block
  98. if current_quality_index > zones_threshold[id_block]:
  99. label_path = os.path.join(label_path, cfg.not_noisy_folder)
  100. else:
  101. label_path = os.path.join(label_path, cfg.noisy_folder)
  102. # Data augmentation!
  103. rotations = [0, 90, 180, 270]
  104. img_flip_labels = ['original', 'horizontal', 'vertical', 'both']
  105. horizontal_img = output_block_img.transpose(Image.FLIP_LEFT_RIGHT)
  106. vertical_img = output_block_img.transpose(Image.FLIP_TOP_BOTTOM)
  107. both_img = output_block_img.transpose(Image.TRANSPOSE)
  108. flip_images = [output_block_img, horizontal_img, vertical_img, both_img]
  109. # rotate and flip image to increase dataset size
  110. for id, flip in enumerate(flip_images):
  111. for rotation in rotations:
  112. rotated_output_img = flip.rotate(rotation)
  113. output_reconstructed_filename = img_path.split('/')[-1].replace('.png', '') + '_' + zones_folder[id_block] + cfg.post_image_name_separator
  114. output_reconstructed_filename = output_reconstructed_filename + img_flip_labels[id] + '_' + str(rotation) + '.png'
  115. output_reconstructed_path = os.path.join(label_path, output_reconstructed_filename)
  116. rotated_output_img.save(output_reconstructed_path)
  117. print(transformation.getName() + "_" + folder_scene + " - " + "{0:.2f}".format(((id_img + 1) / number_scene_image)* 100.) + "%")
  118. sys.stdout.write("\033[F")
  119. print('\n')
  120. print("%s_%s : end of data generation\n" % (transformation.getName(), transformation.getParam()))
  121. def main():
  122. parser = argparse.ArgumentParser(description="Compute and prepare data of feature of all scenes using specific interval if necessary")
  123. parser.add_argument('--features', type=str,
  124. help="list of features choice in order to compute data",
  125. default='svd_reconstruction, ipca_reconstruction',
  126. required=True)
  127. parser.add_argument('--params', type=str,
  128. help="list of specific param for each feature choice (See README.md for further information in 3D mode)",
  129. default='100, 200 :: 50, 25',
  130. required=True)
  131. args = parser.parse_args()
  132. p_features = list(map(str.strip, args.features.split(',')))
  133. p_params = list(map(str.strip, args.params.split('::')))
  134. transformations = []
  135. for id, feature in enumerate(p_features):
  136. if feature not in features_choices:
  137. raise ValueError("Unknown feature, please select a correct feature : ", features_choices)
  138. transformations.append(Transformation(feature, p_params[id]))
  139. # generate all or specific feature data
  140. for transformation in transformations:
  141. generate_data(transformation)
  142. if __name__== "__main__":
  143. main()