ivelin commited on
Commit
97cff20
1 Parent(s): ad1211b

chore: checkpoint

Browse files

Signed-off-by: ivelin <ivelin.eth@gmail.com>

Files changed (1) hide show
  1. ui_refexp.py +24 -12
ui_refexp.py CHANGED
@@ -21,7 +21,7 @@ import os
21
  import tensorflow as tf
22
  import re
23
  import datasets
24
-
25
  import numpy as np
26
 
27
  # Find for instance the citation on arxiv or on the dataset repo/website
@@ -68,7 +68,7 @@ _METADATA_URLS = {
68
  def tfrecord2list(tfr_file: None):
69
  """Filter and convert refexp tfrecord file to a list of dict object.
70
  Each sample in the list is a dict with the following keys: (image_id, prompt, target_bounding_box)"""
71
- test_raw_dataset = tf.data.TFRecordDataset([tfr_file])
72
  count = 0
73
  donut_refexp_dict = []
74
  for raw_record in raw_tfr_dataset:
@@ -141,8 +141,8 @@ class UIRefExp(datasets.GeneratorBasedBuilder):
141
  "screenshot": datasets.Image(),
142
  # click the search button next to menu drawer at the top of the screen
143
  "prompt": datasets.Value("string"),
144
- # [xmin, ymin, xmax, ymax], normalized screen reference values between 0 and 1
145
- "target_bounding_box": dict,
146
  }
147
  )
148
 
@@ -215,13 +215,25 @@ class UIRefExp(datasets.GeneratorBasedBuilder):
215
 
216
  metadata = tfrecord2list(metadata_file)
217
  files_to_keep = set()
 
218
  for sample in metadata:
219
- files_to_keep.add(sample["image_id"])
 
 
 
 
 
 
 
220
  for file_path, file_obj in images:
221
- image_id = file_path.search("(\d+).jpg").group(1)
222
- if image_id and image_id in files_to_keep:
223
- label = file_path.split("/")[2]
224
- yield file_path, {
225
- "image": {"path": file_path, "bytes": file_obj.read()},
226
- "label": label,
227
- }
 
 
 
 
 
21
  import tensorflow as tf
22
  import re
23
  import datasets
24
+ import json
25
  import numpy as np
26
 
27
  # Find for instance the citation on arxiv or on the dataset repo/website
 
68
  def tfrecord2list(tfr_file: None):
69
  """Filter and convert refexp tfrecord file to a list of dict object.
70
  Each sample in the list is a dict with the following keys: (image_id, prompt, target_bounding_box)"""
71
+ raw_tfr_dataset = tf.data.TFRecordDataset([tfr_file])
72
  count = 0
73
  donut_refexp_dict = []
74
  for raw_record in raw_tfr_dataset:
 
141
  "screenshot": datasets.Image(),
142
  # click the search button next to menu drawer at the top of the screen
143
  "prompt": datasets.Value("string"),
144
+ # json: {xmin, ymin, xmax, ymax}, normalized screen reference values between 0 and 1
145
+ "target_bounding_box": datasets.Value("string"),
146
  }
147
  )
148
 
 
215
 
216
  metadata = tfrecord2list(metadata_file)
217
  files_to_keep = set()
218
+ image_labels = {}
219
  for sample in metadata:
220
+ image_id = sample["image_id"]
221
+ files_to_keep.add(image_id)
222
+ labels = image_labels.get(image_id)
223
+ if isinstance(labels, list):
224
+ labels.append(sample)
225
+ else:
226
+ labels = [sample]
227
+ image_labels[image_id] = labels
228
  for file_path, file_obj in images:
229
+ image_id = re.search("(\d+).jpg", file_path)
230
+ if image_id:
231
+ image_id = image_id.group(1)
232
+ if image_id in files_to_keep:
233
+ for labels in image_labels[image_id]:
234
+ bb_json = json.dumps(labels["target_bounding_box"])
235
+ yield file_path, {
236
+ "screenshot": {"path": file_path, "bytes": file_obj.read()},
237
+ "prompt": labels["prompt"],
238
+ "target_bounding_box": bb_json
239
+ }