|
from PIL import Image |
|
from typing import List |
|
from abc import abstractmethod |
|
|
|
from ..utils import InfererModule, ModelWrapper |
|
|
|
class CommonUpscaler(InfererModule): |
|
_VALID_UPSCALE_RATIOS = [] |
|
|
|
async def upscale(self, image_batch: List[Image.Image], upscale_ratio: float) -> List[Image.Image]: |
|
if upscale_ratio == 1: |
|
return image_batch |
|
|
|
self._VALID_UPSCALE_RATIOS.sort() |
|
assert(self._VALID_UPSCALE_RATIOS[0] > 1) |
|
|
|
ratio_left = upscale_ratio |
|
while ratio_left > 0: |
|
ratio = self._VALID_UPSCALE_RATIOS[-1] |
|
for valid_ratio in self._VALID_UPSCALE_RATIOS: |
|
if ratio_left <= valid_ratio: |
|
ratio = valid_ratio |
|
break |
|
ratio_left -= ratio |
|
if upscale_ratio > self._VALID_UPSCALE_RATIOS[-1]: |
|
self.logger.info(f'Upscaling image by {ratio}; left: {ratio_left}') |
|
image_batch = await self._upscale(image_batch, ratio) |
|
if ratio_left < 0: |
|
downscale_ratio = (ratio + ratio_left) / ratio |
|
self.logger.info(f'Downscaling image by {downscale_ratio} to correct upscale ratio') |
|
for i, image in enumerate(image_batch): |
|
image_batch[i] = image.resize((int(image.size[0] * downscale_ratio), int(image.size[1] * downscale_ratio))) |
|
return image_batch |
|
|
|
@abstractmethod |
|
async def _upscale(self, image_batch: List[Image.Image], upscale_ratio: float) -> List[Image.Image]: |
|
pass |
|
|
|
class OfflineUpscaler(CommonUpscaler, ModelWrapper): |
|
_MODEL_SUB_DIR = 'upscaling' |
|
|
|
async def _upscale(self, *args, **kwargs): |
|
return await self.infer(*args, **kwargs) |
|
|
|
@abstractmethod |
|
async def _infer(self, image_batch: List[Image.Image], upscale_ratio: float) -> List[Image.Image]: |
|
""" |
|
Perform the actual upscaling of the images. |
|
|
|
Args: |
|
image_batch: The list of images to upscale. |
|
upscale_ratio: The upscale ratio to use. |
|
|
|
Returns: |
|
The list of upscaled images. |
|
""" |
|
pass |
|
|
