def test_soft_dvpp_decode_random_crop_resize_jpeg(plot=False):
"""
Test SoftDvppDecodeRandomCropResizeJpeg op
"""
logger.info("test_random_decode_resize_op")
# First dataset
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False)
random_crop_decode_resize_op = vision.RandomCropDecodeResize((256, 512), (1, 1), (0.5, 0.5))
data1 = data1.map(operations=random_crop_decode_resize_op, input_columns=["image"])
# Second dataset
data2 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, columns_list=["image"], shuffle=False)
soft_dvpp_random_crop_decode_resize_op = vision.SoftDvppDecodeRandomCropResizeJpeg((256, 512), (1, 1), (0.5, 0.5))
data2 = data2.map(operations=soft_dvpp_random_crop_decode_resize_op, input_columns=["image"])
num_iter = 0
for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1, output_numpy=True),
data2.create_dict_iterator(num_epochs=1, output_numpy=True)):
if num_iter > 0:
break
image1 = item1["image"]
image2 = item2["image"]
mse = diff_mse(image1, image2)
assert mse <= 0.06
logger.info("random_crop_decode_resize_op_{}, mse: {}".format(num_iter + 1, mse))
if plot:
visualize_image(image1, image2, mse)
num_iter += 1
def create_dataset(dataset_path,
do_train,
repeat_num=1,
batch_size=32,
target="GPU",
dtype="fp16",
device_num=1):
ds.config.set_numa_enable(True)
if device_num == 1:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=4,
shuffle=True)
else:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=4,
shuffle=True,
num_shards=device_num,
shard_id=get_rank())
image_size = 224
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
normalize_op = C.Normalize(mean=mean, std=std)
if dtype == "fp16":
if args_opt.eval:
x_dtype = "float32"
else:
x_dtype = "float16"
normalize_op = C.NormalizePad(mean=mean, std=std, dtype=x_dtype)
if do_train:
trans = [
C.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
normalize_op,
]
else:
trans = [
C.Decode(),
C.Resize(256),
C.CenterCrop(image_size),
normalize_op,
]
if dtype == "fp32":
trans.append(C.HWC2CHW())
data_set = data_set.map(operations=trans,
input_columns="image",
num_parallel_workers=8)
# apply batch operations
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
if repeat_num > 1:
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset(dataset_path, config, do_train, repeat_num=1):
"""
create a train or eval dataset
Args:
dataset_path(string): the path of dataset.
config(dict): config of dataset.
do_train(bool): whether dataset is used for train or eval.
repeat_num(int): the repeat times of dataset. Default: 1.
Returns:
dataset
"""
rank = config.rank
group_size = config.group_size
if group_size == 1:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=config.work_nums,
shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=config.work_nums,
shuffle=True,
num_shards=group_size,
shard_id=rank)
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(config.image_size),
C.RandomHorizontalFlip(prob=0.5),
C.RandomColorAdjust(brightness=0.4, saturation=0.5) # fast mode
]
else:
trans = [
C.Decode(),
C.Resize(int(config.image_size / 0.875)),
C.CenterCrop(config.image_size)
]
trans += [
C.Rescale(1.0 / 255.0, 0.0),
C.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans,
input_columns="image",
num_parallel_workers=config.work_nums)
ds = ds.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=config.work_nums)
# apply batch operations
ds = ds.batch(config.batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path, do_train, 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
"""
device_num = int(os.getenv("RANK_SIZE"))
rank_id = int(os.getenv("RANK_ID"))
if device_num == 1:
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=8, shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=8, shuffle=True,
num_shards=device_num, shard_id=rank_id)
image_size = 224
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size, scale=(0.08, 1.0), ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
else:
trans = [
C.Decode(),
C.Resize((256, 256)),
C.CenterCrop(image_size),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans, input_columns="image", num_parallel_workers=8)
ds = ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=8)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path,
do_train,
batch_size=16,
device_num=1,
rank=0):
"""
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.
batch_size(int): the batch size of dataset. Default: 16.
device_num (int): Number of shards that the dataset should be divided into (default=1).
rank (int): The shard ID within num_shards (default=0).
Returns:
dataset
"""
if device_num == 1:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=device_num,
shard_id=rank)
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(299),
C.RandomHorizontalFlip(prob=0.5),
C.RandomColorAdjust(brightness=0.4, contrast=0.4, saturation=0.4)
]
else:
trans = [C.Decode(), C.Resize(320), C.CenterCrop(299)]
trans += [
C.Normalize(mean=[127.5, 127.5, 127.5], std=[127.5, 127.5, 127.5]),
C.HWC2CHW(),
C2.TypeCast(mstype.float32)
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(input_columns="image",
operations=trans,
num_parallel_workers=8)
ds = ds.map(input_columns="label",
operations=type_cast_op,
num_parallel_workers=8)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
return ds
def create_dataset4(dataset_path, do_train, repeat_num=1, batch_size=32, target="Ascend"):
"""
create a train or eval imagenet2012 dataset for se-resnet50
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, rank_id = _get_rank_info()
if device_num == 1:
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=12, shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=12, shuffle=True,
num_shards=device_num, shard_id=rank_id)
image_size = 224
mean = [123.68, 116.78, 103.94]
std = [1.0, 1.0, 1.0]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size, scale=(0.08, 1.0), ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
else:
trans = [
C.Decode(),
C.Resize(292),
C.CenterCrop(256),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans, input_columns="image", num_parallel_workers=12)
ds = ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=12)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path, do_train, repeat_num=1, batch_size=32):
"""
Create a train or eval dataset.
Args:
dataset_path (str): 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.
"""
do_shuffle = bool(do_train)
if device_num == 1 or not do_train:
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=config.work_nums, shuffle=do_shuffle)
else:
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=config.work_nums,
shuffle=do_shuffle, num_shards=device_num, shard_id=device_id)
image_length = 299
if do_train:
trans = [
C.RandomCropDecodeResize(image_length, scale=(0.08, 1.0), ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.RandomColorAdjust(brightness=0.4, contrast=0.4, saturation=0.4)
]
else:
trans = [
C.Decode(),
C.Resize(image_length),
C.CenterCrop(image_length)
]
trans += [
C.Rescale(1.0 / 255.0, 0.0),
C.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(input_columns="label", operations=type_cast_op, num_parallel_workers=config.work_nums)
ds = ds.map(input_columns="image", operations=trans, num_parallel_workers=config.work_nums)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path, do_train, rank, group_size, repeat_num=1):
"""
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.
rank (int): The shard ID within num_shards (default=None).
group_size (int): Number of shards that the dataset should be divided into (default=None).
repeat_num(int): the repeat times of dataset. Default: 1.
Returns:
dataset
"""
if group_size == 1:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=cfg.work_nums,
shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=cfg.work_nums,
shuffle=True,
num_shards=group_size,
shard_id=rank)
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(299,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.RandomColorAdjust(brightness=0.4, contrast=0.4, saturation=0.4)
]
else:
trans = [C.Decode(), C.Resize(299), C.CenterCrop(299)]
trans += [
C.Rescale(1.0 / 255.0, 0.0),
C.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans,
input_columns="image",
num_parallel_workers=cfg.work_nums)
ds = ds.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=cfg.work_nums)
# apply batch operations
ds = ds.batch(cfg.batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path, do_train, device_num=1, rank=0):
"""
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.
rank (int): The shard ID within num_shards (default=None).
group_size (int): Number of shards that the dataset should be divided into (default=None).
repeat_num(int): the repeat times of dataset. Default: 1.
Returns:
dataset
"""
if device_num == 1:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=device_num,
shard_id=rank)
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(224),
C.RandomHorizontalFlip(prob=0.5),
C.RandomColorAdjust(brightness=0.4, contrast=0.4, saturation=0.4)
]
else:
trans = [C.Decode(), C.Resize(239), C.CenterCrop(224)]
trans += [
C.Normalize(mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255]),
C.HWC2CHW(),
]
type_cast_op = C2.TypeCast(mstype.int32)
data_set = data_set.map(input_columns="image",
operations=trans,
num_parallel_workers=8)
data_set = data_set.map(input_columns="label",
operations=type_cast_op,
num_parallel_workers=8)
# apply batch operations
data_set = data_set.batch(config.batch_size, drop_remainder=True)
return data_set
def create_dataset(data_path, is_train=True, batch_size=32):
# import
import mindspore.common.dtype as mstype
import mindspore.dataset.engine as de
import mindspore.dataset.transforms.c_transforms as C2
import mindspore.dataset.vision.c_transforms as C
from mindspore.common import set_seed
set_seed(1)
# shard
num_shards = shard_id = None
rand_size = os.getenv("RANK_SIZE")
rand_id = os.getenv("RANK_ID")
if rand_size is not None and rand_id is not None:
num_shards = int(rand_size)
shard_id = int(rand_id)
# define dataset
data_path = os.path.join(data_path, "train" if is_train else "val")
ds = de.ImageFolderDataset(data_path, shuffle=True, num_parallel_workers=8,
num_shards=num_shards, shard_id=shard_id, num_samples=None)
# define ops
comps_ops = list()
# train or val
if is_train:
comps_ops.append(C.RandomCropDecodeResize(224, scale=(0.08, 1.0), ratio=(0.75, 1.333)))
comps_ops.append(C.RandomHorizontalFlip(prob=0.5))
comps_ops.append(C.RandomColorAdjust(brightness=0.4, contrast=0.4, saturation=0.4))
else:
comps_ops.append(C.Decode())
comps_ops.append(C.Resize(224))
comps_ops.append(C.CenterCrop(224))
comps_ops.append(C.Rescale(1 / 255.0, 0.))
comps_ops.append(C.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))
comps_ops.append(C.HWC2CHW())
# map ops
ds = ds.map(input_columns=["image"], operations=comps_ops, num_parallel_workers=8)
ds = ds.map(input_columns=["label"], operations=C2.TypeCast(mstype.int32), num_parallel_workers=8)
# batch & repeat
ds = ds.batch(batch_size=batch_size, drop_remainder=is_train)
ds = ds.repeat(count=1)
return ds
def create_dataset(dataset_path,
do_train,
repeat_num=1,
batch_size=32,
target="GPU",
dtype="fp16"):
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=4,
shuffle=True)
image_size = 224
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
]
else:
trans = [
C.Decode(),
C.Resize(256),
C.CenterCrop(image_size),
C.Normalize(mean=mean, std=std),
]
if dtype == "fp32":
trans.append(C.HWC2CHW())
ds = ds.map(operations=trans,
input_columns="image",
num_parallel_workers=4)
if dtype == "fp16":
ds = ds.map(operations=pad,
input_columns="image",
num_parallel_workers=4)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
if repeat_num > 1:
ds = ds.repeat(repeat_num)
return ds
def test_random_crop_decode_resize_md5():
"""
Test RandomCropDecodeResize with md5 check
"""
logger.info("Test RandomCropDecodeResize with md5 check")
original_seed = config_get_set_seed(10)
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)
random_crop_decode_resize_op = vision.RandomCropDecodeResize((256, 512), (1, 1), (0.5, 0.5))
data = data.map(operations=random_crop_decode_resize_op, input_columns=["image"])
# Compare with expected md5 from images
filename = "random_crop_decode_resize_01_result.npz"
save_and_check_md5(data, filename, generate_golden=GENERATE_GOLDEN)
# Restore configuration
ds.config.set_seed(original_seed)
ds.config.set_num_parallel_workers((original_num_parallel_workers))
def create_dataset2(dataset_path,
do_train,
repeat_num=1,
batch_size=32,
target="Ascend",
distribute=False):
"""
create a train or eval imagenet2012 dataset for resnet50
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
distribute(bool): data for distribute or not. Default: False
Returns:
dataset
"""
if target == "Ascend":
device_num, rank_id = _get_rank_info()
else:
if distribute:
init()
rank_id = get_rank()
device_num = get_group_size()
else:
device_num = 1
if device_num == 1:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=device_num,
shard_id=rank_id)
image_size = 224
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
else:
trans = [
C.Decode(),
C.Resize(256),
C.CenterCrop(image_size),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
data_set = data_set.map(operations=trans,
input_columns="image",
num_parallel_workers=8)
data_set = data_set.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=8)
# apply batch operations
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset(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"))
columns_list = ['image', 'label']
if config.data_load_mode == "mindrecord":
load_func = partial(de.MindDataset, dataset_path, columns_list)
else:
load_func = partial(de.ImageFolderDataset, dataset_path)
if do_train:
if rank_size == 1:
ds = load_func(num_parallel_workers=8, shuffle=True)
else:
ds = load_func(num_parallel_workers=8,
shuffle=True,
num_shards=rank_size,
shard_id=rank_id)
else:
ds = load_func(num_parallel_workers=8, shuffle=False)
elif device_target == "GPU":
if do_train:
from mindspore.communication.management import get_rank, get_group_size
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=get_group_size(),
shard_id=get_rank())
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
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 = C.Decode()
resize_crop_decode_op = C.RandomCropDecodeResize(resize_height,
scale=(0.08, 1.0),
ratio=(0.75, 1.333))
horizontal_flip_op = C.RandomHorizontalFlip(prob=0.5)
resize_op = C.Resize(256)
center_crop = C.CenterCrop(resize_height)
normalize_op = C.Normalize(mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
change_swap_op = C.HWC2CHW()
if do_train:
trans = [
resize_crop_decode_op, horizontal_flip_op, normalize_op,
change_swap_op
]
else:
trans = [
decode_op, resize_op, center_crop, normalize_op, change_swap_op
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans,
input_columns="image",
num_parallel_workers=16)
ds = ds.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=8)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path, config, buffer_size=1000):
"""
create a train or eval dataset
Args:
dataset_path(string): the path of dataset.
config(struct): the config of train and eval in diffirent platform.
repeat_num(int): the repeat times of dataset. Default: 1.
Returns:
train_dataset, val_dataset
"""
ds = de.ImageFolderDataset(dataset_path, num_parallel_workers=4)
train_ds, eval_ds = ds.split(config.data_split, randomize=True)
resize_height = config.image_height
resize_width = config.image_width
# define operations mapping to each sample
normalize_op = C.Normalize(mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
change_swap_op = C.HWC2CHW()
type_cast_op = C2.TypeCast(mstype.int32)
# operations for training
crop_decode_resize = C.RandomCropDecodeResize(resize_height,
scale=(0.08, 1.0),
ratio=(0.75, 1.333))
horizontal_flip_op = C.RandomHorizontalFlip(prob=0.5)
color_adjust = C.RandomColorAdjust(brightness=0.4,
contrast=0.4,
saturation=0.4)
# operations for inference
decode_op = C.Decode()
resize_op = C.Resize((256, 256))
center_crop = C.CenterCrop(resize_width)
train_trans = [
crop_decode_resize, horizontal_flip_op, color_adjust, normalize_op,
change_swap_op
]
train_ds = train_ds.map(input_columns="image",
operations=train_trans,
num_parallel_workers=4)
train_ds = train_ds.map(input_columns="label",
operations=type_cast_op,
num_parallel_workers=4)
eval_trans = [
decode_op, resize_op, center_crop, normalize_op, change_swap_op
]
eval_ds = eval_ds.map(input_columns="image",
operations=eval_trans,
num_parallel_workers=4)
eval_ds = eval_ds.map(input_columns="label",
operations=type_cast_op,
num_parallel_workers=4)
# apply shuffle operations
train_ds = train_ds.shuffle(buffer_size=buffer_size)
# apply batch operations
train_ds = train_ds.batch(config.batch_size, drop_remainder=True)
eval_ds = eval_ds.batch(config.eval_batch_size, drop_remainder=True)
return train_ds, eval_ds
def create_dataset(dataset_path, do_train, config, repeat_num=1):
"""
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.
config(struct): the config of train and eval in diffirent platform.
repeat_num(int): the repeat times of dataset. Default: 1.
Returns:
dataset
"""
if config.platform == "Ascend":
rank_size = int(os.getenv("RANK_SIZE", '1'))
rank_id = int(os.getenv("RANK_ID", '0'))
if rank_size == 1:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=rank_size,
shard_id=rank_id)
elif config.platform == "GPU":
if do_train:
if config.run_distribute:
from mindspore.communication.management import get_rank, get_group_size
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=get_group_size(),
shard_id=get_rank())
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
elif config.platform == "CPU":
ds = de.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
resize_height = config.image_height
resize_width = config.image_width
buffer_size = 1000
# define map operations
decode_op = C.Decode()
resize_crop_op = C.RandomCropDecodeResize(resize_height,
scale=(0.08, 1.0),
ratio=(0.75, 1.333))
horizontal_flip_op = C.RandomHorizontalFlip(prob=0.5)
resize_op = C.Resize((256, 256))
center_crop = C.CenterCrop(resize_width)
rescale_op = C.RandomColorAdjust(brightness=0.4,
contrast=0.4,
saturation=0.4)
normalize_op = C.Normalize(mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
change_swap_op = C.HWC2CHW()
if do_train:
trans = [
resize_crop_op, horizontal_flip_op, rescale_op, normalize_op,
change_swap_op
]
else:
trans = [
decode_op, resize_op, center_crop, normalize_op, change_swap_op
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans,
input_columns="image",
num_parallel_workers=8)
ds = ds.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=8)
# apply shuffle operations
ds = ds.shuffle(buffer_size=buffer_size)
# apply batch operations
ds = ds.batch(config.batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
def create_dataset(dataset_path,
do_train,
config,
device_target,
repeat_num=1,
batch_size=32,
run_distribute=False):
"""
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 == "GPU":
if do_train:
if run_distribute:
from mindspore.communication.management import get_rank, get_group_size
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=get_group_size(),
shard_id=get_rank())
else:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
raise ValueError("Unsupported device_target.")
resize_height = config.image_height
resize_width = config.image_width
buffer_size = 1000
# define map operations
decode_op = C.Decode()
resize_crop_op = C.RandomCropDecodeResize(resize_height,
scale=(0.08, 1.0),
ratio=(0.75, 1.333))
horizontal_flip_op = C.RandomHorizontalFlip(prob=0.5)
resize_op = C.Resize(256)
center_crop = C.CenterCrop(resize_width)
rescale_op = C.RandomColorAdjust(brightness=0.4,
contrast=0.4,
saturation=0.4)
normalize_op = C.Normalize(mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
change_swap_op = C.HWC2CHW()
if do_train:
trans = [
resize_crop_op, horizontal_flip_op, rescale_op, normalize_op,
change_swap_op
]
else:
trans = [
decode_op, resize_op, center_crop, normalize_op, change_swap_op
]
type_cast_op = C2.TypeCast(mstype.int32)
data_set = data_set.map(operations=trans,
input_columns="image",
num_parallel_workers=8)
data_set = data_set.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=8)
# apply shuffle operations
data_set = data_set.shuffle(buffer_size=buffer_size)
# apply batch operations
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset(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()
rank_id = get_rank()
device_num = get_group_size()
columns_list = ['image', 'label']
if config.data_load_mode == "mindrecord":
load_func = partial(ds.MindDataset, dataset_path, columns_list)
else:
load_func = partial(ds.ImageFolderDataset, dataset_path)
if device_num == 1:
data_set = load_func(num_parallel_workers=8, shuffle=True)
else:
data_set = load_func(num_parallel_workers=8,
shuffle=True,
num_shards=device_num,
shard_id=rank_id)
image_size = 224
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
else:
trans = [
C.Decode(),
C.Resize(256),
C.CenterCrop(image_size),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
data_set = data_set.map(operations=trans,
input_columns="image",
num_parallel_workers=8)
data_set = data_set.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=8)
# apply batch operations
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
data_set = data_set.repeat(repeat_num)
return data_set
def classification_dataset(data_dir,
image_size,
per_batch_size,
max_epoch,
rank,
group_size,
mode='train',
input_mode='folder',
root='',
num_parallel_workers=None,
shuffle=None,
sampler=None,
class_indexing=None,
drop_remainder=True,
transform=None,
target_transform=None):
"""
A function that returns a dataset for classification. The mode of input dataset could be "folder" or "txt".
If it is "folder", all images within one folder have the same label. If it is "txt", all paths of images
are written into a textfile.
Args:
data_dir (str): Path to the root directory that contains the dataset for "input_mode="folder"".
Or path of the textfile that contains every image's path of the dataset.
image_size (Union(int, sequence)): Size of the input images.
per_batch_size (int): the batch size of evey step during training.
max_epoch (int): the number of epochs.
rank (int): The shard ID within num_shards (default=None).
group_size (int): Number of shards that the dataset should be divided
into (default=None).
mode (str): "train" or others. Default: " train".
input_mode (str): The form of the input dataset. "folder" or "txt". Default: "folder".
root (str): the images path for "input_mode="txt"". Default: " ".
num_parallel_workers (int): Number of workers to read the data. Default: None.
shuffle (bool): Whether or not to perform shuffle on the dataset
(default=None, performs shuffle).
sampler (Sampler): Object used to choose samples from the dataset. Default: None.
class_indexing (dict): A str-to-int mapping from folder name to index
(default=None, the folder names will be sorted
alphabetically and each class will be given a
unique index starting from 0).
Examples:
>>> from src.dataset import classification_dataset
>>> # path to imagefolder directory. This directory needs to contain sub-directories which contain the images
>>> data_dir = "/path/to/imagefolder_directory"
>>> de_dataset = classification_dataset(data_dir, image_size=[224, 244],
>>> per_batch_size=64, max_epoch=100,
>>> rank=0, group_size=4)
>>> # Path of the textfile that contains every image's path of the dataset.
>>> data_dir = "/path/to/dataset/images/train.txt"
>>> images_dir = "/path/to/dataset/images"
>>> de_dataset = classification_dataset(data_dir, image_size=[224, 244],
>>> per_batch_size=64, max_epoch=100,
>>> rank=0, group_size=4,
>>> input_mode="txt", root=images_dir)
"""
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
if transform is None:
if mode == 'train':
transform_img = [
V_C.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
V_C.RandomHorizontalFlip(prob=0.5),
V_C.RandomColorAdjust(brightness=0.4,
contrast=0.4,
saturation=0.4),
V_C.Normalize(mean=mean, std=std),
V_C.HWC2CHW()
]
else:
transform_img = [
V_C.Decode(),
V_C.Resize((256, 256)),
V_C.CenterCrop(image_size),
V_C.Normalize(mean=mean, std=std),
V_C.HWC2CHW()
]
else:
transform_img = transform
if target_transform is None:
transform_label = [C.TypeCast(mstype.int32)]
else:
transform_label = target_transform
if input_mode == 'folder':
de_dataset = de.ImageFolderDataset(
data_dir,
num_parallel_workers=num_parallel_workers,
shuffle=shuffle,
sampler=sampler,
class_indexing=class_indexing,
num_shards=group_size,
shard_id=rank)
else:
dataset = TxtDataset(root, data_dir)
sampler = DistributedSampler(dataset,
rank,
group_size,
shuffle=shuffle)
de_dataset = de.GeneratorDataset(dataset, ["image", "label"],
sampler=sampler)
de_dataset = de_dataset.map(operations=transform_img,
input_columns="image",
num_parallel_workers=num_parallel_workers)
de_dataset = de_dataset.map(operations=transform_label,
input_columns="label",
num_parallel_workers=num_parallel_workers)
columns_to_project = ["image", "label"]
de_dataset = de_dataset.project(columns=columns_to_project)
de_dataset = de_dataset.batch(per_batch_size,
drop_remainder=drop_remainder)
de_dataset = de_dataset.repeat(max_epoch)
return de_dataset
def create_dataset_imagenet(dataset_path,
repeat_num=1,
training=True,
num_parallel_workers=None,
shuffle=None):
"""
create a train or eval imagenet2012 dataset for resnet50
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
"""
device_num, rank_id = _get_rank_info()
if device_num == 1:
data_set = ds.ImageFolderDataset(
dataset_path,
num_parallel_workers=num_parallel_workers,
shuffle=shuffle)
else:
data_set = ds.ImageFolderDataset(
dataset_path,
num_parallel_workers=num_parallel_workers,
shuffle=shuffle,
num_shards=device_num,
shard_id=rank_id)
assert imagenet_cfg.image_height == imagenet_cfg.image_width, "image_height not equal image_width"
image_size = imagenet_cfg.image_height
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if training:
transform_img = [
vision.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
vision.RandomHorizontalFlip(prob=0.5),
vision.RandomColorAdjust(0.4, 0.4, 0.4, 0.1),
vision.Normalize(mean=mean, std=std),
vision.HWC2CHW()
]
else:
transform_img = [
vision.Decode(),
vision.Resize(256),
vision.CenterCrop(image_size),
vision.Normalize(mean=mean, std=std),
vision.HWC2CHW()
]
transform_label = [C.TypeCast(mstype.int32)]
data_set = data_set.map(input_columns="image",
num_parallel_workers=12,
operations=transform_img)
data_set = data_set.map(input_columns="label",
num_parallel_workers=4,
operations=transform_label)
# apply batch operations
data_set = data_set.batch(imagenet_cfg.batch_size, drop_remainder=True)
# apply dataset repeat operation
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset4(dataset_path,
do_train,
repeat_num=1,
batch_size=32,
target="Ascend",
distribute=False,
enable_cache=False,
cache_session_id=None):
"""
create a train or eval imagenet2012 dataset for se-resnet50
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
distribute(bool): data for distribute or not. Default: False
enable_cache(bool): whether tensor caching service is used for eval. Default: False
cache_session_id(int): If enable_cache, cache session_id need to be provided. Default: None
Returns:
dataset
"""
if target == "Ascend":
device_num, rank_id = _get_rank_info()
else:
if distribute:
init()
rank_id = get_rank()
device_num = get_group_size()
else:
device_num = 1
if device_num == 1:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=12,
shuffle=True)
else:
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=12,
shuffle=True,
num_shards=device_num,
shard_id=rank_id)
image_size = 224
mean = [123.68, 116.78, 103.94]
std = [1.0, 1.0, 1.0]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
else:
trans = [
C.Decode(),
C.Resize(292),
C.CenterCrop(256),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
data_set = data_set.map(operations=trans,
input_columns="image",
num_parallel_workers=12)
# only enable cache for eval
if do_train:
enable_cache = False
if enable_cache:
if not cache_session_id:
raise ValueError(
"A cache session_id must be provided to use cache.")
eval_cache = ds.DatasetCache(session_id=int(cache_session_id), size=0)
data_set = data_set.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=12,
cache=eval_cache)
else:
data_set = data_set.map(operations=type_cast_op,
input_columns="label",
num_parallel_workers=12)
# apply batch operations
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset(dataset_path,
do_train,
config,
platform,
repeat_num=1,
batch_size=100):
"""
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 platform == "Ascend":
rank_size = int(os.getenv("RANK_SIZE"))
rank_id = int(os.getenv("RANK_ID"))
if rank_size == 1:
data_set = ds.MindDataset(dataset_path,
num_parallel_workers=8,
shuffle=True)
else:
data_set = ds.MindDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=rank_size,
shard_id=rank_id)
elif platform == "GPU":
if do_train:
from mindspore.communication.management import get_rank, get_group_size
data_set = ds.MindDataset(dataset_path,
num_parallel_workers=8,
shuffle=True,
num_shards=get_group_size(),
shard_id=get_rank())
else:
data_set = ds.MindDataset(dataset_path,
num_parallel_workers=8,
shuffle=False)
else:
raise ValueError("Unsupported platform.")
resize_height = config.image_height
buffer_size = 1000
# define map operations
resize_crop_op = C.RandomCropDecodeResize(resize_height,
scale=(0.08, 1.0),
ratio=(0.75, 1.333))
horizontal_flip_op = C.RandomHorizontalFlip(prob=0.5)
color_op = C.RandomColorAdjust(brightness=0.4,
contrast=0.4,
saturation=0.4)
rescale_op = C.Rescale(1 / 255.0, 0)
normalize_op = C.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
change_swap_op = C.HWC2CHW()
# define python operations
decode_p = P.Decode()
resize_p = P.Resize(256, interpolation=Inter.BILINEAR)
center_crop_p = P.CenterCrop(224)
totensor = P.ToTensor()
normalize_p = P.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
composeop = P2.Compose(
[decode_p, resize_p, center_crop_p, totensor, normalize_p])
if do_train:
trans = [
resize_crop_op, horizontal_flip_op, color_op, rescale_op,
normalize_op, change_swap_op
]
else:
trans = composeop
type_cast_op = C2.TypeCast(mstype.int32)
data_set = data_set.map(input_columns="image",
operations=trans,
num_parallel_workers=8)
data_set = data_set.map(input_columns="label_list",
operations=type_cast_op,
num_parallel_workers=8)
# apply shuffle operations
data_set = data_set.shuffle(buffer_size=buffer_size)
# apply batch operations
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset_imagenet(dataset_path,
batch_size=32,
repeat_num=1,
training=True,
num_parallel_workers=None,
shuffle=None,
sampler=None,
class_indexing=None):
"""
create a train or eval imagenet2012 dataset for resnet50
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
"""
device_num, rank_id = _get_rank_info()
cfg = alexnet_imagenet_cfg
num_parallel_workers = 16
if device_num == 1:
num_parallel_workers = 48
ds.config.set_prefetch_size(8)
else:
ds.config.set_numa_enable(True)
data_set = ds.ImageFolderDataset(dataset_path,
num_parallel_workers=4,
shuffle=shuffle,
sampler=sampler,
class_indexing=class_indexing,
num_shards=device_num,
shard_id=rank_id)
assert cfg.image_height == cfg.image_width, "imagenet_cfg.image_height not equal imagenet_cfg.image_width"
image_size = cfg.image_height
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if training:
transform_img = [
CV.RandomCropDecodeResize(image_size,
scale=(0.08, 1.0),
ratio=(0.75, 1.333)),
CV.RandomHorizontalFlip(prob=0.5),
CV.Normalize(mean=mean, std=std),
CV.HWC2CHW()
]
else:
transform_img = [
CV.Decode(),
CV.Resize((256, 256)),
CV.CenterCrop(image_size),
CV.Normalize(mean=mean, std=std),
CV.HWC2CHW()
]
data_set = data_set.map(input_columns="image",
num_parallel_workers=num_parallel_workers,
operations=transform_img)
data_set = data_set.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
if repeat_num > 1:
data_set = data_set.repeat(repeat_num)
return data_set
def create_dataset2(dataset_path, do_train=True, repeat_num=1, batch_size=32, target="gpu", rank=0, size=1):
"""
create a train or eval imagenet2012 dataset for resnet50
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, rank_id = _get_rank_info()
else:
init()
rank_id = rank
device_num = size
file_list = [os.path.join(dataset_path, f'train-{num:05d}-of-01024') for num in range(1024)]
if device_num == 1:
ds = msds.MindDataset(dataset_file=file_list, num_parallel_workers=8, shuffle=True)
else:
ds = msds.MindDataset(dataset_file=file_list, num_parallel_workers=8, shuffle=True,
num_shards=device_num, shard_id=rank_id)
image_size = 224
mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
# define map operations
if do_train:
trans = [
C.RandomCropDecodeResize(image_size, scale=(0.08, 1.0), ratio=(0.75, 1.333)),
C.RandomHorizontalFlip(prob=0.5),
C.Normalize(mean=mean, std=std),
C.HWC2CHW(),
C2.TypeCast(mstype.float16)
]
else:
trans = [
C.Decode(),
C.Resize(256),
C.CenterCrop(image_size),
C.Normalize(mean=mean, std=std),
C.HWC2CHW()
]
type_cast_op = C2.TypeCast(mstype.int32)
ds = ds.map(operations=trans, input_columns="image", num_parallel_workers=8)
ds = ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=8)
# apply batch operations
ds = ds.batch(batch_size, drop_remainder=True)
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
