Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -144,14 +144,23 @@ def process_image(
|
|
144 |
generator.manual_seed(seed)
|
145 |
|
146 |
# Process input image
|
147 |
-
|
148 |
-
|
149 |
|
150 |
-
# Resize
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
155 |
|
156 |
# Move pipeline to GPU for processing
|
157 |
pipeline.to(accelerator.device)
|
@@ -164,12 +173,12 @@ def process_image(
|
|
164 |
False, # tile_diffusion
|
165 |
None, # tile_diffusion_size
|
166 |
None, # tile_diffusion_stride
|
167 |
-
prompt,
|
168 |
-
|
169 |
num_inference_steps=num_inference_steps,
|
170 |
generator=generator,
|
171 |
-
height=
|
172 |
-
width=
|
173 |
guidance_scale=guidance_scale,
|
174 |
negative_prompt=negative_prompt,
|
175 |
conditioning_scale=conditioning_scale,
|
@@ -178,18 +187,21 @@ def process_image(
|
|
178 |
use_vae_encode_condition=False
|
179 |
)
|
180 |
|
181 |
-
|
182 |
-
|
183 |
# Apply color fixing if specified
|
184 |
if color_fix_method != "none":
|
185 |
fix_func = wavelet_color_fix if color_fix_method == "wavelet" else adain_color_fix
|
186 |
-
|
|
|
|
|
|
|
187 |
|
188 |
# Move pipeline back to CPU
|
189 |
pipeline.to("cpu")
|
190 |
torch.cuda.empty_cache()
|
191 |
|
192 |
-
return
|
193 |
|
194 |
except Exception as e:
|
195 |
print(f"Error processing image: {str(e)}")
|
|
|
144 |
generator.manual_seed(seed)
|
145 |
|
146 |
# Process input image
|
147 |
+
validation_image = Image.fromarray(input_image)
|
148 |
+
ori_width, ori_height = validation_image.size
|
149 |
|
150 |
+
# Resize logic from original script
|
151 |
+
resize_flag = False
|
152 |
+
rscale = upscale_factor
|
153 |
+
process_size = 512 # Same as args.process_size in original
|
154 |
+
|
155 |
+
if ori_width < process_size//rscale or ori_height < process_size//rscale:
|
156 |
+
scale = (process_size//rscale)/min(ori_width, ori_height)
|
157 |
+
tmp_image = validation_image.resize((round(scale*ori_width), round(scale*ori_height)))
|
158 |
+
validation_image = tmp_image
|
159 |
+
resize_flag = True
|
160 |
+
|
161 |
+
validation_image = validation_image.resize((validation_image.size[0]*rscale, validation_image.size[1]*rscale))
|
162 |
+
validation_image = validation_image.resize((validation_image.size[0]//8*8, validation_image.size[1]//8*8))
|
163 |
+
width, height = validation_image.size
|
164 |
|
165 |
# Move pipeline to GPU for processing
|
166 |
pipeline.to(accelerator.device)
|
|
|
173 |
False, # tile_diffusion
|
174 |
None, # tile_diffusion_size
|
175 |
None, # tile_diffusion_stride
|
176 |
+
prompt,
|
177 |
+
validation_image,
|
178 |
num_inference_steps=num_inference_steps,
|
179 |
generator=generator,
|
180 |
+
height=height,
|
181 |
+
width=width,
|
182 |
guidance_scale=guidance_scale,
|
183 |
negative_prompt=negative_prompt,
|
184 |
conditioning_scale=conditioning_scale,
|
|
|
187 |
use_vae_encode_condition=False
|
188 |
)
|
189 |
|
190 |
+
image = output.images[0]
|
191 |
+
|
192 |
# Apply color fixing if specified
|
193 |
if color_fix_method != "none":
|
194 |
fix_func = wavelet_color_fix if color_fix_method == "wavelet" else adain_color_fix
|
195 |
+
image = fix_func(image, validation_image)
|
196 |
+
|
197 |
+
if resize_flag:
|
198 |
+
image = image.resize((ori_width*rscale, ori_height*rscale))
|
199 |
|
200 |
# Move pipeline back to CPU
|
201 |
pipeline.to("cpu")
|
202 |
torch.cuda.empty_cache()
|
203 |
|
204 |
+
return image
|
205 |
|
206 |
except Exception as e:
|
207 |
print(f"Error processing image: {str(e)}")
|