def get_dataset(dataset, args):
if dataset.lower() == 'bdd':
train_dataset = gdata.BDDInstance(root='/data1/datasets/bdd100k/',
splits='instances_train2018',
skip_empty=True,
use_color_maps=False)
val_dataset = gdata.BDDInstance(root='/data1/datasets/bdd100k/',
splits='instances_val2018',
skip_empty=False,
use_color_maps=False)
val_metric = BDDInstanceMetric(val_dataset,
args.save_prefix + '_eval',
cleanup=True)
else:
raise NotImplementedError(
'Dataset: {} not implemented.'.format(dataset))
return train_dataset, val_dataset, val_metric
def get_dataset(dataset, args):
if dataset.lower() == 'coco':
train_dataset = gdata.BDDInstance(
root='/Volumes/DATASET/BDD100k/bdd100k/',
splits='bdd100k_to_coco_labels_images_val2018',
skip_empty=True,
use_color_maps=False)
val_dataset = gdata.BDDInstance(
root='/Volumes/DATASET/BDD100k/bdd100k/',
splits='bdd100k_to_coco_labels_images_val2018',
skip_empty=False,
use_color_maps=False)
val_metric = COCOInstanceMetric(val_dataset,
args.save_prefix + '_eval',
cleanup=True)
else:
raise NotImplementedError(
'Dataset: {} not implemented.'.format(dataset))
return train_dataset, val_dataset, val_metric