ShixuanAn commited on
Commit
87e8eff
1 Parent(s): 6eb3c7c

Update RDD_2020.py

Browse files
Files changed (1) hide show
  1. RDD_2020.py +47 -41
RDD_2020.py CHANGED
@@ -57,8 +57,8 @@ class RDD2020_Dataset(datasets.GeneratorBasedBuilder):
57
  "height": datasets.Value("int32"),
58
  "depth": datasets.Value("int32"),
59
  }),
60
- "image": datasets.Value("string"),
61
- #"image_path": datasets.Value("string"),
62
  #"pics_array": datasets.Array3D(shape=(None, None, 3), dtype="uint8"),
63
  "crack_type": datasets.Sequence(datasets.Value("string")),
64
  "crack_coordinates": datasets.Sequence(datasets.Features({
@@ -116,50 +116,56 @@ class RDD2020_Dataset(datasets.GeneratorBasedBuilder):
116
 
117
 
118
  def _generate_examples(self, filepath, split):
 
 
119
  for country_dir in ['Czech', 'India', 'Japan']:
120
  images_dir = f"{filepath}/{country_dir}/images"
 
 
 
 
 
121
  annotations_dir = f"{filepath}/{country_dir}/annotations/xmls" if split == "train" else None
122
-
 
123
  for image_file in os.listdir(images_dir):
124
  if not image_file.endswith('.jpg'):
125
  continue
126
-
127
  image_id = f"{image_file.split('.')[0]}"
 
128
  image_path = os.path.join(images_dir, image_file)
129
-
130
- # Load the image as a PIL Image object
131
- with Image.open(image_path) as img:
132
- # If you need to process the image, do it here
133
-
134
- if annotations_dir:
135
- annotation_file = image_id + '.xml'
136
- annotation_path = os.path.join(annotations_dir, annotation_file)
137
- if not os.path.exists(annotation_path):
138
- continue
139
- tree = ET.parse(annotation_path)
140
- root = tree.getroot()
141
- crack_type = []
142
- crack_coordinates = []
143
- for obj in root.findall('object'):
144
- crack_type.append(obj.find('name').text)
145
- bndbox = obj.find('bndbox')
146
- coordinates = {
147
- "x_min": int(bndbox.find('xmin').text),
148
- "x_max": int(bndbox.find('xmax').text),
149
- "y_min": int(bndbox.find('ymin').text),
150
- "y_max": int(bndbox.find('ymax').text),
151
- }
152
- crack_coordinates.append(coordinates)
153
- else:
154
- crack_type = []
155
- crack_coordinates = []
156
-
157
- # Now 'img' is a PIL Image object, we keep it open and yield it
158
- yield image_id, {
159
- "image_id": image_id,
160
- "country": country_dir,
161
- "type": split,
162
- "image": img.copy(), # We make a copy of the image object
163
- "crack_type": crack_type,
164
- "crack_coordinates": crack_coordinates,
165
- }
 
57
  "height": datasets.Value("int32"),
58
  "depth": datasets.Value("int32"),
59
  }),
60
+ # "image": datasets.Value("string"),
61
+ "image_path": datasets.Value("string"),
62
  #"pics_array": datasets.Array3D(shape=(None, None, 3), dtype="uint8"),
63
  "crack_type": datasets.Sequence(datasets.Value("string")),
64
  "crack_coordinates": datasets.Sequence(datasets.Features({
 
116
 
117
 
118
  def _generate_examples(self, filepath, split):
119
+
120
+ # Iterate over each country directory
121
  for country_dir in ['Czech', 'India', 'Japan']:
122
  images_dir = f"{filepath}/{country_dir}/images"
123
+
124
+ # print(os.listdir(filepath))
125
+ # print(os.listdir(filepath))
126
+ # print(os.listdir(f"{filepath}/{country_dir}"))
127
+
128
  annotations_dir = f"{filepath}/{country_dir}/annotations/xmls" if split == "train" else None
129
+
130
+ # Iterate over each image in the country's image directory
131
  for image_file in os.listdir(images_dir):
132
  if not image_file.endswith('.jpg'):
133
  continue
134
+
135
  image_id = f"{image_file.split('.')[0]}"
136
+
137
  image_path = os.path.join(images_dir, image_file)
138
+ # img = Image.open(image_path)
139
+
140
+ if annotations_dir:
141
+ annotation_file = image_id + '.xml'
142
+ annotation_path = os.path.join(annotations_dir, annotation_file)
143
+ if not os.path.exists(annotation_path):
144
+ continue
145
+ tree = ET.parse(annotation_path)
146
+ root = tree.getroot()
147
+ crack_type = []
148
+ crack_coordinates = []
149
+ for obj in root.findall('object'):
150
+ crack_type.append(obj.find('name').text)
151
+ bndbox = obj.find('bndbox')
152
+ coordinates = {
153
+ "x_min": int(bndbox.find('xmin').text),
154
+ "x_max": int(bndbox.find('xmax').text),
155
+ "y_min": int(bndbox.find('ymin').text),
156
+ "y_max": int(bndbox.find('ymax').text),
157
+ }
158
+ crack_coordinates.append(coordinates)
159
+ else:
160
+ crack_type = []
161
+ crack_coordinates = []
162
+
163
+ yield image_id, {
164
+ "image_id": image_id,
165
+ "country": country_dir,
166
+ "type": split,
167
+ "image_path": image_path,
168
+ # "image": img,
169
+ "crack_type": crack_type,
170
+ "crack_coordinates": crack_coordinates,
171
+ }