vkashko commited on
Commit
b744c0b
·
1 Parent(s): bdc95f7

feat: upload script

Browse files
face_masks.py → facial_keypoint_detection.py RENAMED
@@ -93,7 +93,8 @@ class FacialKeypointDetection(datasets.GeneratorBasedBuilder):
93
  features=datasets.Features({
94
  'image_id': datasets.Value('uint32'),
95
  'image': datasets.Image(),
96
- 'mask': datasets.Image()
 
97
  }),
98
  supervised_keys=None,
99
  homepage=_HOMEPAGE,
@@ -105,44 +106,38 @@ class FacialKeypointDetection(datasets.GeneratorBasedBuilder):
105
  masks = dl_manager.download_and_extract(f"{_DATA}masks.zip")
106
  annotations = dl_manager.download(f"{_DATA}{_NAME}.csv")
107
  images = dl_manager.iter_files(images)
 
 
108
  return [
109
  datasets.SplitGenerator(name=datasets.Split.TRAIN,
110
  gen_kwargs={
111
  "images": images,
 
112
  'annotations': annotations
113
  }),
114
  ]
115
 
116
- def _generate_examples(self, images, annotations):
117
  annotations_df = pd.read_csv(annotations, sep=',')
118
- images_data = pd.DataFrame(columns=['Link', 'Path'])
119
- for idx, image_path in enumerate(images):
 
120
  images_data.loc[idx] = {
121
- 'Link': '/'.join(image_path.split('/')[-2:]),
122
- 'Path': image_path
 
123
  }
124
 
125
  annotations_df = pd.merge(annotations_df,
126
  images_data,
127
  how='left',
128
- on=['Link'])
129
- for idx, worker_id in enumerate(pd.unique(annotations_df['WorkerId'])):
130
- annotation: pd.DataFrame = annotations_df.loc[
131
- annotations_df['WorkerId'] == worker_id]
132
- annotation = annotation.sort_values(['Link'])
133
- data = {
134
- f'photo_{row[5]}': load_image_file(row[7])
135
- for row in annotation.itertuples()
 
136
  }
137
-
138
- age = annotation.loc[annotation['Type'] == 1]['Age'].values[0]
139
- country = annotation.loc[annotation['Type'] ==
140
- 1]['Country'].values[0]
141
- sex = annotation.loc[annotation['Type'] == 1]['Sex'].values[0]
142
-
143
- data['worker_id'] = worker_id
144
- data['age'] = age
145
- data['country'] = country
146
- data['sex'] = sex
147
-
148
- yield idx, data
 
93
  features=datasets.Features({
94
  'image_id': datasets.Value('uint32'),
95
  'image': datasets.Image(),
96
+ 'mask': datasets.Image(),
97
+ 'key_points': datasets.Value('string')
98
  }),
99
  supervised_keys=None,
100
  homepage=_HOMEPAGE,
 
106
  masks = dl_manager.download_and_extract(f"{_DATA}masks.zip")
107
  annotations = dl_manager.download(f"{_DATA}{_NAME}.csv")
108
  images = dl_manager.iter_files(images)
109
+ masks = dl_manager.iter_files(masks)
110
+
111
  return [
112
  datasets.SplitGenerator(name=datasets.Split.TRAIN,
113
  gen_kwargs={
114
  "images": images,
115
+ "masks": masks,
116
  'annotations': annotations
117
  }),
118
  ]
119
 
120
+ def _generate_examples(self, images, masks, annotations):
121
  annotations_df = pd.read_csv(annotations, sep=',')
122
+ images_data = pd.DataFrame(
123
+ columns=['image_name', 'image_path', 'mask_path'])
124
+ for idx, (image_path, mask_path) in enumerate(zip(images, masks)):
125
  images_data.loc[idx] = {
126
+ 'image_name': image_path.split('/')[-1],
127
+ 'image_path': image_path,
128
+ 'mask_path': mask_path
129
  }
130
 
131
  annotations_df = pd.merge(annotations_df,
132
  images_data,
133
  how='left',
134
+ on=['image_name'])
135
+
136
+ for row in annotations_df.sort_values(['image_name'
137
+ ]).itertuples(index=False):
138
+ yield idx, {
139
+ 'image_id': row[0],
140
+ 'image': load_image_file(row[3]),
141
+ 'mask': load_image_file(row[4]),
142
+ 'key_points': row[2]
143
  }