Syoy commited on
Commit
5b1a6d7
1 Parent(s): a68df04

modified dataset building script to load embeddings as List[Value(float)] -> ds.to_pandas() compatible

Browse files
Files changed (1) hide show
  1. dcase23-task2-enriched.py +5 -18
dcase23-task2-enriched.py CHANGED
@@ -298,15 +298,6 @@ class DCASE2023Task2DatasetConfig(datasets.BuilderConfig):
298
  raise NotImplementedError
299
 
300
  if type(data) == datasets.Dataset:
301
- # remove embedding columns first -> throws error in .to_pandas()
302
- embeddings = {}
303
- emb_features = [key for key, val in data.features.items() if type(val) == datasets.Array2D]
304
- if len(emb_features) > 0:
305
- embeddings = {
306
- key: [np.asarray(emb).reshape(-1,) for emb in data[key].copy()] for key in emb_features
307
- }
308
- data = data.remove_columns(emb_features)
309
-
310
  # retrieve split
311
  df = data.to_pandas()
312
  df["split"] = data.split._name if "+" not in data.split._name else df["path"].map(get_split)
@@ -315,10 +306,6 @@ class DCASE2023Task2DatasetConfig(datasets.BuilderConfig):
315
  # get clearnames for classes
316
  class_names = data.features["class"].names
317
  df["class_name"] = df["class"].apply(lambda x: class_names[x])
318
-
319
- # append embeddings
320
- for emb_name, emb_list in embeddings.items():
321
- df[emb_name] = emb_list
322
  elif type(data) == pd.DataFrame:
323
  df = data
324
  else:
@@ -341,7 +328,7 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
341
  """Dataset for the DCASE 2023 Challenge Task 2 "First-Shot Unsupervised Anomalous Sound Detection
342
  for Machine Condition Monitoring"."""
343
 
344
- VERSION = datasets.Version("0.0.3")
345
 
346
  DEFAULT_CONFIG_NAME = "dev"
347
 
@@ -365,7 +352,7 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
365
  features = {
366
  "audio": datasets.Audio(sampling_rate=16_000),
367
  "path": datasets.Value("string"),
368
- "section": datasets.Value("int64"),
369
  "domain": datasets.ClassLabel(num_classes=2, names=["source", "target"]),
370
  "label": datasets.ClassLabel(num_classes=_NUM_TARGETS, names=_TARGET_NAMES),
371
  "class": datasets.ClassLabel(num_classes=_NUM_CLASSES, names=_CLASS_NAMES),
@@ -378,7 +365,7 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
378
  }
379
  if self.config.embeddings_urls is not None:
380
  features.update({
381
- emb_name: datasets.Array2D(shape=emb["size"], dtype=emb["dtype"]) for emb_name, emb in self.config.embeddings_urls.items()
382
  })
383
  features = datasets.Features(features)
384
 
@@ -408,7 +395,7 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
408
  audio_path[split] = dl_manager.download(self.config.data_urls[split])
409
  local_extracted_archive[split] = dl_manager.extract(
410
  audio_path[split]) if not dl_manager.is_streaming else None
411
- if self.config.embeddings_urls is not None and not dl_manager.is_streaming:
412
  for emb_name, emb_data in self.config.embeddings_urls.items():
413
  downloaded_embeddings = dl_manager.download(emb_data[split])
414
  embeddings[split][emb_name] = np.load(downloaded_embeddings, allow_pickle=True)["arr_0"].item()
@@ -447,7 +434,7 @@ class DCASE2023Task2Dataset(datasets.GeneratorBasedBuilder):
447
  result = {field: None for field in data_fields}
448
  result.update(metadata[metadata["path"] == lookup].T.squeeze().to_dict())
449
  for emb_key in embeddings.keys():
450
- result[emb_key] = embeddings[emb_key][lookup]
451
  result["path"] = path
452
  yield id_, {**result, "audio": audio}
453
  id_ += 1
 
298
  raise NotImplementedError
299
 
300
  if type(data) == datasets.Dataset:
 
 
 
 
 
 
 
 
 
301
  # retrieve split
302
  df = data.to_pandas()
303
  df["split"] = data.split._name if "+" not in data.split._name else df["path"].map(get_split)
 
306
  # get clearnames for classes
307
  class_names = data.features["class"].names
308
  df["class_name"] = df["class"].apply(lambda x: class_names[x])
 
 
 
 
309
  elif type(data) == pd.DataFrame:
310
  df = data
311
  else:
 
328
  """Dataset for the DCASE 2023 Challenge Task 2 "First-Shot Unsupervised Anomalous Sound Detection
329
  for Machine Condition Monitoring"."""
330
 
331
+ VERSION = datasets.Version("0.0.4")
332
 
333
  DEFAULT_CONFIG_NAME = "dev"
334
 
 
352
  features = {
353
  "audio": datasets.Audio(sampling_rate=16_000),
354
  "path": datasets.Value("string"),
355
+ "section": datasets.Value("uint32"),
356
  "domain": datasets.ClassLabel(num_classes=2, names=["source", "target"]),
357
  "label": datasets.ClassLabel(num_classes=_NUM_TARGETS, names=_TARGET_NAMES),
358
  "class": datasets.ClassLabel(num_classes=_NUM_CLASSES, names=_CLASS_NAMES),
 
365
  }
366
  if self.config.embeddings_urls is not None:
367
  features.update({
368
+ emb_name: [datasets.Value(emb["dtype"])] for emb_name, emb in self.config.embeddings_urls.items()
369
  })
370
  features = datasets.Features(features)
371
 
 
395
  audio_path[split] = dl_manager.download(self.config.data_urls[split])
396
  local_extracted_archive[split] = dl_manager.extract(
397
  audio_path[split]) if not dl_manager.is_streaming else None
398
+ if self.config.embeddings_urls is not None:
399
  for emb_name, emb_data in self.config.embeddings_urls.items():
400
  downloaded_embeddings = dl_manager.download(emb_data[split])
401
  embeddings[split][emb_name] = np.load(downloaded_embeddings, allow_pickle=True)["arr_0"].item()
 
434
  result = {field: None for field in data_fields}
435
  result.update(metadata[metadata["path"] == lookup].T.squeeze().to_dict())
436
  for emb_key in embeddings.keys():
437
+ result[emb_key] = np.asarray(embeddings[emb_key][lookup]).squeeze().tolist()
438
  result["path"] = path
439
  yield id_, {**result, "audio": audio}
440
  id_ += 1