openvinotoolkit/stable-diffusion-webui
1
1
Fork 0
mirror of https://github.com/openvinotoolkit/stable-diffusion-webui.git synced 2024-12-15 23:22:48 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fixes LDSR upscaler producing black bars

Browse Source
This commit is contained in:
Han Lin 2022-11-06 11:08:20 +08:00
parent 2f47724b73
commit 6603f63b7b
1 changed files with 11 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
modules/ldsr_model_arch.py
View File

@ -101,8 +101,8 @@ class LDSR:
down_sample_rate = target_scale / 4 down_sample_rate = target_scale / 4
wd = width_og * down_sample_rate wd = width_og * down_sample_rate
hd = height_og * down_sample_rate hd = height_og * down_sample_rate
width_downsampled_pre = int(wd) width_downsampled_pre = int(np.ceil(wd))
height_downsampled_pre = int(hd) height_downsampled_pre = int(np.ceil(hd))
if down_sample_rate != 1: if down_sample_rate != 1:
print( print(
@ -110,7 +110,12 @@ class LDSR:
im_og = im_og.resize((width_downsampled_pre, height_downsampled_pre), Image.LANCZOS) im_og = im_og.resize((width_downsampled_pre, height_downsampled_pre), Image.LANCZOS)
else: else:
print(f"Down sample rate is 1 from {target_scale} / 4 (Not downsampling)") print(f"Down sample rate is 1 from {target_scale} / 4 (Not downsampling)")
logs = self.run(model["model"], im_og, diffusion_steps, eta)
# pad width and height to multiples of 64, pads with the edge values of image to avoid artifacts
pad_w, pad_h = np.max(((2, 2), np.ceil(np.array(im_og.size) / 64).astype(int)), axis=0) * 64 - im_og.size
im_padded = Image.fromarray(np.pad(np.array(im_og), ((0, pad_h), (0, pad_w), (0, 0)), mode='edge'))
logs = self.run(model["model"], im_padded, diffusion_steps, eta)
sample = logs["sample"] sample = logs["sample"]
sample = sample.detach().cpu() sample = sample.detach().cpu()
@ -120,6 +125,9 @@ class LDSR:
sample = np.transpose(sample, (0, 2, 3, 1)) sample = np.transpose(sample, (0, 2, 3, 1))
a = Image.fromarray(sample[0]) a = Image.fromarray(sample[0])
# remove padding
a = a.crop((0, 0) + tuple(np.array(im_og.size) * 4))
del model del model
gc.collect() gc.collect()
torch.cuda.empty_cache() torch.cuda.empty_cache()