SushantGautam commited on
Commit
ad713e3
·
verified ·
1 Parent(s): c58c4fe

Update kvasir-points_datasets_script.py

Browse files
Files changed (1) hide show
  1. kvasir-points_datasets_script.py +40 -17
kvasir-points_datasets_script.py CHANGED
@@ -15,6 +15,7 @@
15
  """Dataset for filtered Kvasir-instrument and Hyper-Kvasir with bounding boxes."""
16
 
17
  import os
 
18
  import json
19
  from PIL import Image
20
  import datasets
@@ -22,17 +23,14 @@ import datasets
22
  import os
23
  import json
24
  import pandas as pd
25
- import hashlib
26
  from collections import defaultdict
27
  import numpy as np
28
 
29
 
30
- def cal_mid(bx): return [[[float(box['xmin'] + box['xmax']) / 2,
31
- float(box['ymin'] + box['ymax']) / 2] for box in bx]]
32
 
33
-
34
- def cal_mid_xy(bx): return [{"x": float(box['xmin'] + box['xmax']) / 2,
35
- "y": float(box['ymin'] + box['ymax']) / 2} for box in bx]
36
 
37
 
38
  def cal_sha256(file_path): return hashlib.sha256(
@@ -51,6 +49,23 @@ def convert_to_json_format(file_path, image_width, image_height):
51
  }
52
  for line in file.readlines()
53
  ]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
54
 
55
 
56
  class_map = {"0": "normal", "1": "cluster", "2": "pinhead"}
@@ -111,7 +126,9 @@ class KvasirHyperBBox(datasets.GeneratorBasedBuilder):
111
  features = datasets.Features({
112
  "image_data": datasets.Image(),
113
  "image_sha256": datasets.Value("string"),
114
- "points": datasets.Sequence(datasets.Sequence(datasets.Sequence(datasets.Value("float32")))),
 
 
115
  "count": datasets.Value("int64"),
116
  "label": datasets.Value("string"),
117
  "collection_method": datasets.Value("string"),
@@ -138,11 +155,14 @@ class KvasirHyperBBox(datasets.GeneratorBasedBuilder):
138
  def _generate_examples(self):
139
  for key, entry in hyper_seg_imgs.items():
140
  img_path = os.path.join(hyper_seg_img_base_path, f"{key}.jpg")
 
141
  hyper_entry = hyper_df.loc[hyper_df['Video file'] == key].iloc[0]
142
  yield key, {
143
- "image_data": open(img_path, 'rb').read(),
144
  "image_sha256": cal_sha256(img_path),
145
- "points": cal_mid(entry['bbox']),
 
 
146
  "count": len(entry['bbox']),
147
  "label": hyper_entry.Finding,
148
  "collection_method": 'counting',
@@ -152,10 +172,13 @@ class KvasirHyperBBox(datasets.GeneratorBasedBuilder):
152
 
153
  for key, entry in instr_seg_imgs.items():
154
  img_path = os.path.join(instr_seg_img_base_path, f"{key}.jpg")
 
155
  yield key, {
156
- "image_data": open(img_path, 'rb').read(),
157
  "image_sha256": cal_sha256(img_path),
158
- "points": cal_mid(entry['bbox']),
 
 
159
  "count": len(entry['bbox']),
160
  "label": "instrument",
161
  "collection_method": "counting",
@@ -166,24 +189,24 @@ class KvasirHyperBBox(datasets.GeneratorBasedBuilder):
166
  for folder in os.listdir(visem_root):
167
  folder_path = os.path.join(visem_root, folder)
168
  labels_all = os.listdir(folder_path+"/labels")
169
- images = os.listdir(folder_path+"/images")
170
- height, width = Image.open(os.path.join(
171
- folder_path, "images", images[0])).size
172
  labels = [labels_all[i] for i in np.linspace(
173
  0, len(labels_all)-1, 250).astype(int)]
174
  for label in labels:
175
  label_path = os.path.join(folder_path, "labels", label)
176
  image_path = label_path.replace(
177
  "/labels/", "/images/").replace(".txt", ".jpg")
 
178
  entry_bbox = convert_to_json_format(label_path, width, height)
179
  label_dict = defaultdict(list)
180
  for entry in entry_bbox:
181
  label_dict[entry['label']].append(entry)
182
  for label in label_dict:
183
  yield cal_sha256(image_path)+label, {
184
- "image_data": open(image_path, 'rb').read(),
185
  "image_sha256": cal_sha256(image_path),
186
- "points": cal_mid(label_dict[label]),
 
 
187
  "count": len(label_dict[label]),
188
  "label": class_map[label],
189
  "collection_method": "counting",
@@ -193,5 +216,5 @@ class KvasirHyperBBox(datasets.GeneratorBasedBuilder):
193
 
194
 
195
  # rm -rf /home/sushant/.cache/huggingface/modules/datasets_modules/datasets/kvasir-points_datasets_script/ /home/sushant/.cache/huggingface/datasets/kvasir-points_datasets_script
196
- # datasets-cli test /global/D1/projects/HOST/Datasets/hyper-kvasir/sushant-experiments/kvasir-points_datasets_script.py --save_info --all_configs --trust_remote_cod
197
  # huggingface-cli upload kvasir-points . . --repo-type dataset
 
15
  """Dataset for filtered Kvasir-instrument and Hyper-Kvasir with bounding boxes."""
16
 
17
  import os
18
+ import io
19
  import json
20
  from PIL import Image
21
  import datasets
 
23
  import os
24
  import json
25
  import pandas as pd
26
+ import hashlib
27
  from collections import defaultdict
28
  import numpy as np
29
 
30
 
 
 
31
 
32
+ # def cal_mid_xy(bx): return [{"x": float(box['xmin'] + box['xmax']) / 2,
33
+ # "y": float(box['ymin'] + box['ymax']) / 2} for box in bx]
 
34
 
35
 
36
  def cal_sha256(file_path): return hashlib.sha256(
 
49
  }
50
  for line in file.readlines()
51
  ]
52
+
53
+ def get_image_bytes(img_path, max_width=700):
54
+ img = Image.open(img_path)
55
+ if img.width <= max_width:
56
+ return open(img_path, "rb").read(), img.width, img.height, 1.0
57
+ with io.BytesIO() as b:
58
+ img.resize((max_width, int(img.height * max_width / img.width))).save(b, "PNG")
59
+ return b.getvalue(), max_width, int(img.height * max_width / img.width), float(max_width) / img.width
60
+
61
+
62
+ def get_bboxes(bx, ratio):
63
+ return [[box[k] * ratio for k in ('xmin', 'ymin', 'xmax', 'ymax')] for box in bx]
64
+
65
+
66
+ def cal_mid(bx, ratio):
67
+ return [[((box['xmin'] + box['xmax']) / 2) * ratio,
68
+ ((box['ymin'] + box['ymax']) / 2) * ratio] for box in bx]
69
 
70
 
71
  class_map = {"0": "normal", "1": "cluster", "2": "pinhead"}
 
126
  features = datasets.Features({
127
  "image_data": datasets.Image(),
128
  "image_sha256": datasets.Value("string"),
129
+ "img_size": datasets.Sequence(datasets.Value("float32")),
130
+ "points": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
131
+ "bbox": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
132
  "count": datasets.Value("int64"),
133
  "label": datasets.Value("string"),
134
  "collection_method": datasets.Value("string"),
 
155
  def _generate_examples(self):
156
  for key, entry in hyper_seg_imgs.items():
157
  img_path = os.path.join(hyper_seg_img_base_path, f"{key}.jpg")
158
+ img, width, height, ratio = get_image_bytes(img_path)
159
  hyper_entry = hyper_df.loc[hyper_df['Video file'] == key].iloc[0]
160
  yield key, {
161
+ "image_data": img,
162
  "image_sha256": cal_sha256(img_path),
163
+ "img_size": [width, height],
164
+ "points": cal_mid(entry['bbox'], ratio),
165
+ "bbox": get_bboxes(entry['bbox'], ratio),
166
  "count": len(entry['bbox']),
167
  "label": hyper_entry.Finding,
168
  "collection_method": 'counting',
 
172
 
173
  for key, entry in instr_seg_imgs.items():
174
  img_path = os.path.join(instr_seg_img_base_path, f"{key}.jpg")
175
+ img, width, height, ratio = get_image_bytes(img_path)
176
  yield key, {
177
+ "image_data": img,
178
  "image_sha256": cal_sha256(img_path),
179
+ "img_size": [width, height],
180
+ "points": cal_mid(entry['bbox'], ratio),
181
+ "bbox": get_bboxes(entry['bbox'], ratio),
182
  "count": len(entry['bbox']),
183
  "label": "instrument",
184
  "collection_method": "counting",
 
189
  for folder in os.listdir(visem_root):
190
  folder_path = os.path.join(visem_root, folder)
191
  labels_all = os.listdir(folder_path+"/labels")
 
 
 
192
  labels = [labels_all[i] for i in np.linspace(
193
  0, len(labels_all)-1, 250).astype(int)]
194
  for label in labels:
195
  label_path = os.path.join(folder_path, "labels", label)
196
  image_path = label_path.replace(
197
  "/labels/", "/images/").replace(".txt", ".jpg")
198
+ img, width, height, ratio = get_image_bytes(image_path)
199
  entry_bbox = convert_to_json_format(label_path, width, height)
200
  label_dict = defaultdict(list)
201
  for entry in entry_bbox:
202
  label_dict[entry['label']].append(entry)
203
  for label in label_dict:
204
  yield cal_sha256(image_path)+label, {
205
+ "image_data": img,
206
  "image_sha256": cal_sha256(image_path),
207
+ "img_size": [width, height],
208
+ "points": cal_mid(label_dict[label], ratio),
209
+ "bbox": get_bboxes(label_dict[label], ratio),
210
  "count": len(label_dict[label]),
211
  "label": class_map[label],
212
  "collection_method": "counting",
 
216
 
217
 
218
  # rm -rf /home/sushant/.cache/huggingface/modules/datasets_modules/datasets/kvasir-points_datasets_script/ /home/sushant/.cache/huggingface/datasets/kvasir-points_datasets_script
219
+ # datasets-cli test /global/D1/projects/HOST/Datasets/hyper-kvasir/sushant-experiments/kvasir-points_datasets_script.py --save_info --all_configs --trust_remote_code
220
  # huggingface-cli upload kvasir-points . . --repo-type dataset