def test_random_horizontal_invalid_prob_py():
"""
Test RandomHorizontalFlip op in py_transforms: invalid input, expect to raise error
"""
logger.info("test_random_horizontal_invalid_prob_py")
# Generate dataset
data = ds.TFRecordDataset(DATA_DIR,
SCHEMA_DIR,
columns_list=["image"],
shuffle=False)
try:
transforms = [
py_vision.Decode(),
# Note: Valid range of prob should be [0.0, 1.0]
py_vision.RandomHorizontalFlip(1.5),
py_vision.ToTensor()
]
transform = py_vision.ComposeOp(transforms)
data = data.map(input_columns=["image"], operations=transform())
except ValueError as e:
logger.info("Got an exception in DE: {}".format(str(e)))
assert "Input prob is not within the required interval of (0.0 to 1.0)." in str(
e)
def test_random_horizontal_valid_prob_py():
"""
Test RandomHorizontalFlip op with py_transforms: valid non-default input, expect to pass
"""
logger.info("test_random_horizontal_valid_prob_py")
original_seed = config_get_set_seed(0)
original_num_parallel_workers = config_get_set_num_parallel_workers(1)
# Generate dataset
data = ds.TFRecordDataset(DATA_DIR,
SCHEMA_DIR,
columns_list=["image"],
shuffle=False)
transforms = [
py_vision.Decode(),
py_vision.RandomHorizontalFlip(0.8),
py_vision.ToTensor()
]
transform = py_vision.ComposeOp(transforms)
data = data.map(input_columns=["image"], operations=transform())
filename = "random_horizontal_01_py_result.npz"
save_and_check_md5(data, filename, generate_golden=GENERATE_GOLDEN)
# Restore config setting
ds.config.set_seed(original_seed)
ds.config.set_num_parallel_workers(original_num_parallel_workers)
def create_dataset_py(dataset_path, do_train, config, device_target, repeat_num=1, batch_size=32):
"""
create a train or eval dataset
Args:
dataset_path(string): the path of dataset.
do_train(bool): whether dataset is used for train or eval.
repeat_num(int): the repeat times of dataset. Default: 1.
batch_size(int): the batch size of dataset. Default: 32.
Returns:
dataset
"""
if device_target == "Ascend":
rank_size = int(os.getenv("RANK_SIZE"))
rank_id = int(os.getenv("RANK_ID"))
if do_train:
if rank_size == 1:
ds = de.ImageFolderDatasetV2(dataset_path, num_parallel_workers=8, shuffle=True)
else:
ds = de.ImageFolderDatasetV2(dataset_path, num_parallel_workers=8, shuffle=True,
num_shards=rank_size, shard_id=rank_id)
else:
ds = de.ImageFolderDatasetV2(dataset_path, num_parallel_workers=8, shuffle=False)
else:
raise ValueError("Unsupported device target.")
resize_height = config.image_height
if do_train:
buffer_size = 20480
# apply shuffle operations
ds = ds.shuffle(buffer_size=buffer_size)
# define map operations
decode_op = P.Decode()
resize_crop_op = P.RandomResizedCrop(resize_height, scale=(0.08, 1.0), ratio=(0.75, 1.333))
horizontal_flip_op = P.RandomHorizontalFlip(prob=0.5)
resize_op = P.Resize(256)
center_crop = P.CenterCrop(resize_height)
to_tensor = P.ToTensor()
normalize_op = P.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
if do_train:
trans = [decode_op, resize_crop_op, horizontal_flip_op, to_tensor, normalize_op]
else:
trans = [decode_op, resize_op, center_crop, to_tensor, normalize_op]
compose = P.ComposeOp(trans)
ds = ds.map(input_columns="image", operations=compose(), num_parallel_workers=8, python_multiprocessing=True)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_imagenet_dataset(imagenet_dir):
ds = de.ImageFolderDatasetV2(imagenet_dir)
transform = F.ComposeOp([
F.Decode(),
F.RandomHorizontalFlip(0.5),
F.ToTensor(),
F.Normalize((0.491, 0.482, 0.447), (0.247, 0.243, 0.262)),
F.RandomErasing()
])
ds = ds.map(input_columns="image", operations=transform())
ds = ds.shuffle(buffer_size=5)
ds = ds.repeat(3)
return ds
def test_random_horizontal_comp(plot=False):
"""
Test test_random_horizontal_flip and compare between python and c image augmentation ops
"""
logger.info("test_random_horizontal_comp")
# First dataset
data1 = ds.TFRecordDataset(DATA_DIR,
SCHEMA_DIR,
columns_list=["image"],
shuffle=False)
decode_op = c_vision.Decode()
# Note: The image must be flipped if prob is set to be 1
random_horizontal_op = c_vision.RandomHorizontalFlip(1)
data1 = data1.map(input_columns=["image"], operations=decode_op)
data1 = data1.map(input_columns=["image"], operations=random_horizontal_op)
# Second dataset
data2 = ds.TFRecordDataset(DATA_DIR,
SCHEMA_DIR,
columns_list=["image"],
shuffle=False)
transforms = [
py_vision.Decode(),
# Note: The image must be flipped if prob is set to be 1
py_vision.RandomHorizontalFlip(1),
py_vision.ToTensor()
]
transform = py_vision.ComposeOp(transforms)
data2 = data2.map(input_columns=["image"], operations=transform())
images_list_c = []
images_list_py = []
for item1, item2 in zip(data1.create_dict_iterator(),
data2.create_dict_iterator()):
image_c = item1["image"]
image_py = (item2["image"].transpose(1, 2, 0) * 255).astype(np.uint8)
images_list_c.append(image_c)
images_list_py.append(image_py)
# Check if the output images are the same
mse = diff_mse(image_c, image_py)
assert mse < 0.001
if plot:
visualize_list(images_list_c, images_list_py, visualize_mode=2)
def create_dataset_py(dataset_path,
do_train,
repeat_num=1,
batch_size=32,
target="Ascend"):
"""
create a train or eval dataset
Args:
dataset_path(string): the path of dataset.
do_train(bool): whether dataset is used for train or eval.
repeat_num(int): the repeat times of dataset. Default: 1
batch_size(int): the batch size of dataset. Default: 32
target(str): the device target. Default: Ascend
Returns:
dataset
"""
if target == "Ascend":
device_num = int(os.getenv("RANK_SIZE"))
rank_id = int(os.getenv("RANK_ID"))
else:
init("nccl")
rank_id = get_rank()
device_num = get_group_size()
if do_train:
if device_num == 1:
ds = de.ImageFolderDatasetV2(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
ds = de.ImageFolderDatasetV2(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=device_num,
shard_id=rank_id)
else:
ds = de.ImageFolderDatasetV2(dataset_path,
num_parallel_workers=8,
shuffle=False)
image_size = 224
# define map operations
decode_op = P.Decode()
resize_crop_op = P.RandomResizedCrop(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333))
horizontal_flip_op = P.RandomHorizontalFlip(prob=0.5)
resize_op = P.Resize(256)
center_crop = P.CenterCrop(image_size)
to_tensor = P.ToTensor()
normalize_op = P.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
# define map operations
if do_train:
trans = [
decode_op, resize_crop_op, horizontal_flip_op, to_tensor,
normalize_op
]
else:
trans = [decode_op, resize_op, center_crop, to_tensor, normalize_op]
compose = P.ComposeOp(trans)
ds = ds.map(input_columns="image",
operations=compose(),
num_parallel_workers=8,
python_multiprocessing=True)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds