130 lines
4.2 KiB
Python
130 lines
4.2 KiB
Python
from logging import getLogger
|
|
from typing import Callable, Optional, Tuple
|
|
|
|
import numpy as np
|
|
import torch
|
|
from PIL import Image
|
|
|
|
from ..diffusers.load import load_pipeline
|
|
from ..diffusers.utils import get_latents_from_seed
|
|
from ..image import expand_image, mask_filter_none, noise_source_histogram
|
|
from ..output import save_image
|
|
from ..params import Border, ImageParams, Size, SizeChart, StageParams
|
|
from ..server import ServerContext
|
|
from ..utils import is_debug
|
|
from ..worker import ProgressCallback, WorkerContext
|
|
from .tile import process_tile_order
|
|
|
|
logger = getLogger(__name__)
|
|
|
|
|
|
class BlendInpaintStage:
|
|
def run(
|
|
self,
|
|
job: WorkerContext,
|
|
server: ServerContext,
|
|
stage: StageParams,
|
|
params: ImageParams,
|
|
source: Image.Image,
|
|
*,
|
|
expand: Border,
|
|
stage_source: Optional[Image.Image] = None,
|
|
stage_mask: Optional[Image.Image] = None,
|
|
fill_color: str = "white",
|
|
mask_filter: Callable = mask_filter_none,
|
|
noise_source: Callable = noise_source_histogram,
|
|
callback: Optional[ProgressCallback] = None,
|
|
**kwargs,
|
|
) -> Image.Image:
|
|
params = params.with_args(**kwargs)
|
|
expand = expand.with_args(**kwargs)
|
|
source = source or stage_source
|
|
logger.info(
|
|
"blending image using inpaint, %s steps: %s", params.steps, params.prompt
|
|
)
|
|
|
|
if stage_mask is None:
|
|
# if no mask was provided, keep the full source image
|
|
stage_mask = Image.new("RGB", source.size, "black")
|
|
|
|
source, stage_mask, noise, _full_dims = expand_image(
|
|
source,
|
|
stage_mask,
|
|
expand,
|
|
fill=fill_color,
|
|
noise_source=noise_source,
|
|
mask_filter=mask_filter,
|
|
)
|
|
|
|
if is_debug():
|
|
save_image(server, "last-source.png", source)
|
|
save_image(server, "last-mask.png", stage_mask)
|
|
save_image(server, "last-noise.png", noise)
|
|
|
|
pipe_type = "lpw" if params.lpw() else "inpaint"
|
|
pipe = load_pipeline(
|
|
server,
|
|
params,
|
|
pipe_type,
|
|
job.get_device(),
|
|
# TODO: add LoRAs and TIs
|
|
)
|
|
|
|
def outpaint(tile_source: Image.Image, dims: Tuple[int, int, int]):
|
|
left, top, tile = dims
|
|
size = Size(*tile_source.size)
|
|
tile_mask = stage_mask.crop((left, top, left + tile, top + tile))
|
|
|
|
if is_debug():
|
|
save_image(server, "tile-source.png", tile_source)
|
|
save_image(server, "tile-mask.png", tile_mask)
|
|
|
|
latents = get_latents_from_seed(params.seed, size)
|
|
if params.lpw():
|
|
logger.debug("using LPW pipeline for inpaint")
|
|
rng = torch.manual_seed(params.seed)
|
|
result = pipe.inpaint(
|
|
params.prompt,
|
|
generator=rng,
|
|
guidance_scale=params.cfg,
|
|
height=size.height,
|
|
image=tile_source,
|
|
latents=latents,
|
|
mask_image=tile_mask,
|
|
negative_prompt=params.negative_prompt,
|
|
num_inference_steps=params.steps,
|
|
width=size.width,
|
|
eta=params.eta,
|
|
callback=callback,
|
|
)
|
|
else:
|
|
rng = np.random.RandomState(params.seed)
|
|
result = pipe(
|
|
params.prompt,
|
|
generator=rng,
|
|
guidance_scale=params.cfg,
|
|
height=size.height,
|
|
image=tile_source,
|
|
latents=latents,
|
|
mask_image=stage_mask,
|
|
negative_prompt=params.negative_prompt,
|
|
num_inference_steps=params.steps,
|
|
width=size.width,
|
|
eta=params.eta,
|
|
callback=callback,
|
|
)
|
|
|
|
return result.images[0]
|
|
|
|
output = process_tile_order(
|
|
stage.tile_order,
|
|
source,
|
|
SizeChart.auto,
|
|
1,
|
|
[outpaint],
|
|
overlap=params.overlap,
|
|
)
|
|
|
|
logger.info("final output image size: %s", output.size)
|
|
return output
|