Datasets:

Modalities:
Image
Formats:
parquet
Languages:
English
DOI:
Libraries:
Datasets
Dask
License:
jpodivin commited on
Commit
ff4d367
1 Parent(s): 0035722

Improved mask conversion script

Browse files

Signed-off-by: Jiri Podivin <jpodivin@gmail.com>

Files changed (1) hide show
  1. utils/convert_masks.py +31 -5
utils/convert_masks.py CHANGED
@@ -1,5 +1,8 @@
1
  #!/bin/env python
2
 
 
 
 
3
  import numpy as np
4
  import json
5
  import os
@@ -13,6 +16,8 @@ import os
13
  import tempfile
14
  import argparse
15
  import glob
 
 
16
 
17
  class InputStream:
18
  def __init__(self, data):
@@ -174,25 +179,46 @@ def merge_masks(mask_metadata, target_mask_dir, label2id):
174
 
175
  def main():
176
  parser = argparse.ArgumentParser('maskconvert')
177
- parser.add_argument('dataset_root')
178
 
179
  arguments = parser.parse_args()
180
- annotations_folder_path = os.path.join(arguments.dataset_root, 'labels_raw')
181
- tmp_mask_path = tempfile.mkdtemp('masks')
 
 
 
 
 
 
 
182
 
183
  files_to_process = glob.glob(f"{annotations_folder_path}/*")
184
 
185
- # For sanity check
186
- source_files = [os.path.basename(name) for name in glob.glob(f"sourcedata/**/*.jpg")]
 
 
 
187
 
188
  metadata = pd.DataFrame(process_files_in_parallel(files_to_process, tmp_mask_path, source_files))
189
 
 
190
  id2label = {int(k): v for k, v in enumerate(['void', 'Fruit', 'Leaf', 'Flower', 'Stem'])}
191
 
192
  label2id = {v: k for k, v in id2label.items()}
193
 
194
  result = merge_masks(metadata, os.path.join(arguments.dataset_root, 'semantic_masks'), label2id)
 
195
  result = pd.DataFrame(result).drop_duplicates()
 
 
 
 
 
 
 
 
 
196
  result.to_csv(
197
  os.path.join(arguments.dataset_root, 'semantic_metadata.csv'),
198
  index=False)
 
1
  #!/bin/env python
2
 
3
+ # Mask conversion for plantorgans
4
+ # 2024 by Jiri Podivin
5
+
6
  import numpy as np
7
  import json
8
  import os
 
16
  import tempfile
17
  import argparse
18
  import glob
19
+ import shutil
20
+ import tarfile
21
 
22
  class InputStream:
23
  def __init__(self, data):
 
179
 
180
  def main():
181
  parser = argparse.ArgumentParser('maskconvert')
182
+ parser.add_argument('dataset_root', help="Root directory of the dataset repo.")
183
 
184
  arguments = parser.parse_args()
185
+
186
+ # Unzip raw labels to temporary location
187
+ tmp_raw_label_pth = tempfile.mkdtemp('_labels_raw')
188
+ with tarfile.open(os.path.join(arguments.dataset_root, 'labels_raw.tar.gz'), mode='r:gz') as archive:
189
+ archive.extractall(tmp_raw_label_pth)
190
+
191
+ annotations_folder_path = os.path.join(tmp_raw_label_pth, 'labels_raw')
192
+
193
+ tmp_mask_path = tempfile.mkdtemp('_masks')
194
 
195
  files_to_process = glob.glob(f"{annotations_folder_path}/*")
196
 
197
+ # Image name -> path dict
198
+ images = {os.path.basename(x): x for x in glob.glob('sourcedata/**/*.jpg')}
199
+
200
+ # Image names
201
+ source_files = [k for k in images.keys()]
202
 
203
  metadata = pd.DataFrame(process_files_in_parallel(files_to_process, tmp_mask_path, source_files))
204
 
205
+ # Label map order IS important
206
  id2label = {int(k): v for k, v in enumerate(['void', 'Fruit', 'Leaf', 'Flower', 'Stem'])}
207
 
208
  label2id = {v: k for k, v in id2label.items()}
209
 
210
  result = merge_masks(metadata, os.path.join(arguments.dataset_root, 'semantic_masks'), label2id)
211
+
212
  result = pd.DataFrame(result).drop_duplicates()
213
+
214
+ # Moving newly labeled images to right dir
215
+ images = {os.path.basename(x): x for x in glob.glob('sourcedata/**/*.jpg')}
216
+
217
+ for img in metadata['image']:
218
+ if 'unlabeled' in images[img]:
219
+ print("Moving {img} to labeled!")
220
+ shutil.move(images[img], os.path.join('sourcedata/labeled/', img))
221
+
222
  result.to_csv(
223
  os.path.join(arguments.dataset_root, 'semantic_metadata.csv'),
224
  index=False)