lisawen commited on
Commit
321149c
1 Parent(s): 009b97c

Update soybean_dataset.py

Browse files
Files changed (1) hide show
  1. soybean_dataset.py +34 -16
soybean_dataset.py CHANGED
@@ -236,45 +236,63 @@ class SoybeanDataset(datasets.GeneratorBasedBuilder):
236
  name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["valid"]}),
237
  ]
238
 
239
- def process_image(self,image_url):
 
240
  response = requests.get(image_url)
241
- response.raise_for_status() # This will raise an exception if there is a download error
242
-
243
- # Open the image from the downloaded bytes and return the PIL Image
244
  img = Image.open(BytesIO(response.content))
245
  return img
246
-
247
-
 
 
 
 
 
 
 
 
 
 
 
 
 
 
248
 
249
  def _generate_examples(self, filepath):
250
- #"""Yields examples as (key, example) tuples."""
251
  logging.info("generating examples from = %s", filepath)
252
 
 
253
  with open(filepath, encoding="utf-8") as f:
254
  data = csv.DictReader(f)
 
 
 
255
 
 
 
256
 
 
 
257
  for row in data:
258
- # Assuming the 'original_image' column has the full path to the image file
259
  unique_id = row['unique_id']
260
  original_image_path = row['original_image']
261
  segmentation_image_path = row['segmentation_image']
262
  sets = row['sets']
263
 
264
- original_image = self.process_image(original_image_path)
265
- segmentation_image = self.process_image(segmentation_image_path)
 
 
 
 
266
 
267
-
268
- # Here you need to replace 'initial_radius', 'final_radius', 'initial_angle', 'final_angle', 'target'
269
- # with actual columns from your CSV or additional processing you need to do
270
- yield row['unique_id'], {
271
  "unique_id": unique_id,
272
  "sets": sets,
273
  "original_image": original_image,
274
  "segmentation_image": segmentation_image,
275
- # ... add other features if necessary
276
  }
277
-
278
 
279
 
280
 
 
236
  name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["valid"]}),
237
  ]
238
 
239
+ def download_image(self, image_url):
240
+ try:
241
  response = requests.get(image_url)
242
+ response.raise_for_status() # This will raise an exception for HTTP errors
 
 
243
  img = Image.open(BytesIO(response.content))
244
  return img
245
+ except requests.RequestException as e:
246
+ logging.error(f"Error downloading {image_url}: {e}")
247
+ return None
248
+
249
+ def download_images_concurrently(self, image_urls):
250
+ images = {}
251
+ with ThreadPoolExecutor(max_workers=5) as executor:
252
+ future_to_url = {executor.submit(self.download_image, url): url for url in image_urls}
253
+ for future in as_completed(future_to_url):
254
+ url = future_to_url[future]
255
+ try:
256
+ image = future.result()
257
+ images[url] = image
258
+ except Exception as exc:
259
+ logging.error(f'{url} generated an exception: {exc}')
260
+ return images
261
 
262
  def _generate_examples(self, filepath):
 
263
  logging.info("generating examples from = %s", filepath)
264
 
265
+ image_urls = []
266
  with open(filepath, encoding="utf-8") as f:
267
  data = csv.DictReader(f)
268
+ for row in data:
269
+ image_urls.append(row['original_image'])
270
+ image_urls.append(row['segmentation_image'])
271
 
272
+ # Download all images concurrently
273
+ downloaded_images = self.download_images_concurrently(set(image_urls)) # Use set to avoid duplicate downloads
274
 
275
+ with open(filepath, encoding="utf-8") as f:
276
+ data = csv.DictReader(f)
277
  for row in data:
 
278
  unique_id = row['unique_id']
279
  original_image_path = row['original_image']
280
  segmentation_image_path = row['segmentation_image']
281
  sets = row['sets']
282
 
283
+ original_image = downloaded_images.get(original_image_path)
284
+ segmentation_image = downloaded_images.get(segmentation_image_path)
285
+
286
+ if original_image is None or segmentation_image is None:
287
+ logging.error(f"Missing image for {unique_id}")
288
+ continue
289
 
290
+ yield unique_id, {
 
 
 
291
  "unique_id": unique_id,
292
  "sets": sets,
293
  "original_image": original_image,
294
  "segmentation_image": segmentation_image,
 
295
  }
 
296
 
297
 
298