# Copyright 2022 ByteDance Ltd. and/or its affiliates.
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
from typing import Tuple, Union, Sequence, Dict, Any, List
import sys
matx = sys.modules['matx']
from .. import ASYNC, RESIZE_NOT_SMALLER, RESIZE_DEFAULT
from .. import INTER_NEAREST, INTER_LINEAR, INTER_CUBIC, INTER_LANCZOS4
from .. import ResizeOp, RandomResizedCropOp
from ._base import BaseInterfaceClass, BatchBaseClass
from ._common import _setup_size, _torch_interp_mode, create_device
[文档]class Resize(BaseInterfaceClass):
[文档] def __init__(self,
size: List[int],
interpolation: str = "bilinear",
resize_mode: str = "",
max_size: int = 0,
device_id: int = -2,
sync: int = ASYNC) -> None:
super().__init__(device_id=device_id, sync=sync)
# check size
if len(size) == 1:
self._size: Tuple[int, int] = (size[0], size[0])
if not resize_mode:
self._resize_mode: str = RESIZE_NOT_SMALLER
else:
self._resize_mode: str = resize_mode
elif len(size) == 2:
self._size: Tuple[int, int] = (size[0], size[1])
if not resize_mode:
self._resize_mode: str = RESIZE_DEFAULT
else:
self._resize_mode: str = resize_mode
else:
assert False, "Resize size value should be an integer or a list/tuple with length 2."
self._max_size: int = max_size
# check interpolation mode
assert interpolation in ["nearest", "bilinear", "bicubic",
"lanczos"], "interpolation_mode should be nearest, bilinear, bicubic or lanczos."
torch_interp_mode: Dict[str, str] = {
"nearest": INTER_NEAREST,
"bilinear": INTER_LINEAR,
"bicubic": INTER_CUBIC,
"lanczos": INTER_LANCZOS4
}
self._interpolation_mode: str = torch_interp_mode[interpolation]
[文档] def __call__(self, device: Any, device_str: str, sync: int) -> Any:
return ResizeImpl(
device,
device_str,
self._size,
self._max_size,
self._resize_mode,
self._interpolation_mode,
sync)
class ResizeImpl(BatchBaseClass):
def __init__(self,
device: Any,
device_str: str,
size: Tuple[int, int],
max_size: int,
resize_mode: str,
interpolation_mode: str,
sync: int = ASYNC) -> None:
super().__init__()
self.device_str: str = device_str
self.size: Tuple[int, int] = size
self.max_size: int = max_size
self.resize_mode: str = resize_mode
self.interpolation_mode: str = interpolation_mode
self.sync: int = sync
self.op: ResizeOp = ResizeOp(
device,
self.size,
self.max_size,
self.interpolation_mode,
self.resize_mode)
self.name: str = "ResizeImpl"
def _process_resize_op(self, imgs: List[matx.NDArray]) -> List[matx.NDArray]:
imgs = self.op(imgs, sync=self.sync)
return imgs
def _process(self, imgs: List[matx.NDArray]) -> List[matx.NDArray]:
return self._process_resize_op(imgs)
def __repr__(self) -> str:
return self.name + '(size={0}, interpolation={1}, max_size={2}, device={3}, sync={4})'.format(
self.size, self.interpolation_mode, self.max_size, self.device_str, self.sync)
[文档]class RandomResizedCrop(BaseInterfaceClass):
[文档] def __init__(self,
size: List[int],
scale: Tuple[float, float] = (0.08, 1.0),
ratio: Tuple[float, float] = (3. / 4., 4. / 3.),
interpolation: str = "bilinear",
device_id: int = -2,
sync: int = ASYNC) -> None:
super().__init__(device_id=device_id, sync=sync)
# check size
if len(size) == 1:
self._size: Tuple[int, int] = (size[0], size[0])
elif len(size) == 2:
self._size: Tuple[int, int] = (size[0], size[1])
else:
assert False, "Resize size value should be an integer or a list/tuple with length 2."
# check scale
if len(scale) == 2:
assert scale[0] < scale[1], "Scale should be of kind [min, max]."
self._scale: List[float] = [scale[0], scale[1]]
else:
assert False, "Scale should be a tuple with length 2."
# check ratio
if len(ratio) == 2:
assert ratio[0] < ratio[1], "Ratio should be of kind [min, max]."
self._ratio: List[float] = [ratio[0], ratio[1]]
else:
assert False, "Ratio should be a tuple with length 2."
# check interpolation mode
assert interpolation in ["nearest", "bilinear", "bicubic",
"lanczos"], "interpolation_mode should be nearest, bilinear, bicubic or lanczos."
torch_interp_mode: Dict[str, str] = {
"nearest": INTER_NEAREST,
"bilinear": INTER_LINEAR,
"bicubic": INTER_CUBIC,
"lanczos": INTER_LANCZOS4
}
self._interpolation_mode: str = torch_interp_mode[interpolation]
[文档] def __call__(self, device: Any, device_str: str, sync: int) -> Any:
return RandomResizedCropImpl(
device,
device_str,
self._size,
self._scale,
self._ratio,
self._interpolation_mode,
sync)
class RandomResizedCropImpl(BatchBaseClass):
def __init__(self,
device: Any,
device_str: str,
size: Tuple[int, int],
scale: List[float],
ratio: List[float],
interpolation_mode: str,
sync: int) -> None:
self.device_str: str = device_str
self.size: Tuple[int, int] = size
self.scale: List[float] = scale
self.ratio: List[float] = ratio
self.interpolation_mode: str = interpolation_mode
self.sync: int = sync
self.op: RandomResizedCropOp = RandomResizedCropOp(
device, self.size, self.scale, self.ratio, self.interpolation_mode)
self.name: str = "RandomResizedCropImpl"
def _process_random_resized_crop_op(self, imgs: List[matx.NDArray]) -> List[matx.NDArray]:
imgs = self.op(imgs, sync=self.sync)
return imgs
def _process(self, imgs: List[matx.NDArray]) -> List[matx.NDArray]:
return self._process_random_resized_crop_op(imgs)
def __repr__(self) -> str:
return self.name + '(size={0}, scale={1}, ratio={2}, interpolation={3}, device={4}, sync={5})'.format(
self.size, self.scale, self.ratio, self.interpolation_mode, self.device_str, self.sync)