def test_concat_ade(separate_eval):
test_dataset = ADE20KDataset(
pipeline=[],
img_dir=osp.join(osp.dirname(__file__), '../data/pseudo_dataset/imgs'))
assert len(test_dataset) == 5
concat_dataset = ConcatDataset([test_dataset, test_dataset],
separate_eval=separate_eval)
assert len(concat_dataset) == 10
# Test format_results
pseudo_results = []
for _ in range(len(concat_dataset)):
h, w = (2, 2)
pseudo_results.append(np.random.randint(low=0, high=7, size=(h, w)))
# test format per image
file_paths = []
for i in range(len(pseudo_results)):
file_paths.extend(
concat_dataset.format_results([pseudo_results[i]],
'.format_ade',
indices=[i]))
assert len(file_paths) == len(concat_dataset)
temp = np.array(Image.open(file_paths[0]))
assert np.allclose(temp, pseudo_results[0] + 1)
shutil.rmtree('.format_ade')
# test default argument
file_paths = concat_dataset.format_results(pseudo_results, '.format_ade')
assert len(file_paths) == len(concat_dataset)
temp = np.array(Image.open(file_paths[0]))
assert np.allclose(temp, pseudo_results[0] + 1)
shutil.rmtree('.format_ade')
def ade20k_datasets(FLAGS):
img_norm_cfg = dict(mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True)
crop_size = (512, 512)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=True),
dict(type='Resize', img_scale=(2048, 512), ratio_range=(0.5, 2.0)),
dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='PhotoMetricDistortion'),
dict(type='Normalize', **img_norm_cfg),
dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_semantic_seg']),
]
test_pipeline = [
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(2048, 512),
# img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75],
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(type='Normalize', **img_norm_cfg),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img']),
])
]
train_pipeline = Compose(train_pipeline)
val_pipeline = Compose(test_pipeline)
train_set = ADE20KDataset(data_root=FLAGS.data_root,
img_dir='images/training',
ann_dir='annotations/training',
pipeline=train_pipeline)
val_set = ADE20KDataset(data_root=FLAGS.data_root,
img_dir='images/validation',
ann_dir='annotations/validation',
pipeline=val_pipeline,
test_mode=True)
return train_set, val_set, None
def test_ade():
test_dataset = ADE20KDataset(
pipeline=[],
img_dir=osp.join(osp.dirname(__file__), '../data/pseudo_dataset/imgs'))
assert len(test_dataset) == 5
# Test format_results
pseudo_results = []
for _ in range(len(test_dataset)):
h, w = (2, 2)
pseudo_results.append(np.random.randint(low=0, high=7, size=(h, w)))
file_paths = test_dataset.format_results(pseudo_results, '.format_ade')
assert len(file_paths) == len(test_dataset)
temp = np.array(Image.open(file_paths[0]))
assert np.allclose(temp, pseudo_results[0] + 1)
shutil.rmtree('.format_ade')
def test_concat_cityscapes(separate_eval):
cityscape_dataset = CityscapesDataset(
pipeline=[],
img_dir=osp.join(osp.dirname(__file__),
'../data/pseudo_cityscapes_dataset/leftImg8bit'),
ann_dir=osp.join(osp.dirname(__file__),
'../data/pseudo_cityscapes_dataset/gtFine'))
assert len(cityscape_dataset) == 1
with pytest.raises(NotImplementedError):
_ = ConcatDataset([cityscape_dataset, cityscape_dataset],
separate_eval=separate_eval)
ade_dataset = ADE20KDataset(pipeline=[],
img_dir=osp.join(
osp.dirname(__file__),
'../data/pseudo_dataset/imgs'))
assert len(ade_dataset) == 5
with pytest.raises(NotImplementedError):
_ = ConcatDataset([cityscape_dataset, ade_dataset],
separate_eval=separate_eval)