|
|
@ -77,12 +77,12 @@ class Imagenet22k(tfds.core.GeneratorBasedBuilder):
|
|
|
|
'validation': self._generate_examples(val_records, manual_dir),
|
|
|
|
'validation': self._generate_examples(val_records, manual_dir),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
def _generate_examples(self, records, manual_dir, alt_label=None, resize_short=True, max_img_size=MAX_DIM):
|
|
|
|
def _generate_examples(self, records, manual_dir, resize_short=True, max_img_size=MAX_DIM):
|
|
|
|
"""Yields examples."""
|
|
|
|
"""Yields examples."""
|
|
|
|
for r in records:
|
|
|
|
for r in records:
|
|
|
|
try:
|
|
|
|
try:
|
|
|
|
filename, output_record = _process_record(
|
|
|
|
filename, output_record = _process_record(
|
|
|
|
r, manual_dir, alt_label=alt_label, resize_short=resize_short, max_img_size=max_img_size)
|
|
|
|
r, manual_dir, resize_short=resize_short, max_img_size=max_img_size)
|
|
|
|
yield filename, output_record
|
|
|
|
yield filename, output_record
|
|
|
|
except Exception as e:
|
|
|
|
except Exception as e:
|
|
|
|
print('Exception:', e)
|
|
|
|
print('Exception:', e)
|
|
|
@ -114,8 +114,6 @@ def _load_records(
|
|
|
|
train_csv,
|
|
|
|
train_csv,
|
|
|
|
validation_csv,
|
|
|
|
validation_csv,
|
|
|
|
labels,
|
|
|
|
labels,
|
|
|
|
alt_labels=None,
|
|
|
|
|
|
|
|
alt_label_name='',
|
|
|
|
|
|
|
|
min_img_size=MIN_DIM,
|
|
|
|
min_img_size=MIN_DIM,
|
|
|
|
):
|
|
|
|
):
|
|
|
|
pd = tfds.core.lazy_imports.pandas
|
|
|
|
pd = tfds.core.lazy_imports.pandas
|
|
|
@ -133,12 +131,10 @@ def _load_records(
|
|
|
|
|
|
|
|
|
|
|
|
train_record_df['label'] = train_record_df['cls'].map(class_to_idx).astype(int)
|
|
|
|
train_record_df['label'] = train_record_df['cls'].map(class_to_idx).astype(int)
|
|
|
|
train_record_df = train_record_df[['filename', 'label']]
|
|
|
|
train_record_df = train_record_df[['filename', 'label']]
|
|
|
|
train_record_df = train_record_df.sample(frac=1, random_state=42)
|
|
|
|
|
|
|
|
print('num train records:', len(train_record_df.index))
|
|
|
|
print('num train records:', len(train_record_df.index))
|
|
|
|
|
|
|
|
|
|
|
|
val_record_df['label'] = val_record_df['cls'].map(class_to_idx).astype(int)
|
|
|
|
val_record_df['label'] = val_record_df['cls'].map(class_to_idx).astype(int)
|
|
|
|
val_record_df = val_record_df[['filename', 'label']]
|
|
|
|
val_record_df = val_record_df[['filename', 'label']]
|
|
|
|
val_record_df = val_record_df.sample(frac=1, random_state=42)
|
|
|
|
|
|
|
|
print('num val records:', len(val_record_df.index))
|
|
|
|
print('num val records:', len(val_record_df.index))
|
|
|
|
|
|
|
|
|
|
|
|
train_records = train_record_df.to_records(index=False)
|
|
|
|
train_records = train_record_df.to_records(index=False)
|
|
|
|