Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -180,10 +180,71 @@ except Exception as e:
|
|
180 |
# --- UI Constants and Helpers ---
|
181 |
MAX_SEED = np.iinfo(np.int32).max
|
182 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
183 |
# --- Main Inference Function ---
|
184 |
@spaces.GPU(duration=60)
|
185 |
def infer(
|
186 |
-
|
187 |
prompt,
|
188 |
seed=42,
|
189 |
randomize_seed=False,
|
@@ -203,6 +264,20 @@ def infer(
|
|
203 |
|
204 |
# Set up the generator for reproducibility
|
205 |
generator = torch.Generator(device=device).manual_seed(seed)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
206 |
|
207 |
print(f"Original prompt: '{prompt}'")
|
208 |
print(f"Negative Prompt: '{negative_prompt}'")
|
@@ -215,7 +290,7 @@ def infer(
|
|
215 |
# Generate the edited image - always generate just 1 image
|
216 |
try:
|
217 |
images = pipe(
|
218 |
-
|
219 |
prompt=prompt,
|
220 |
negative_prompt=negative_prompt,
|
221 |
num_inference_steps=num_inference_steps,
|
@@ -268,10 +343,16 @@ with gr.Blocks(css=css) as demo:
|
|
268 |
|
269 |
with gr.Row():
|
270 |
with gr.Column():
|
271 |
-
|
272 |
-
label="
|
273 |
-
show_label=True,
|
274 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
275 |
)
|
276 |
# Changed from Gallery to Image
|
277 |
result = gr.Image(
|
@@ -329,7 +410,7 @@ with gr.Blocks(css=css) as demo:
|
|
329 |
triggers=[run_button.click, prompt.submit],
|
330 |
fn=infer,
|
331 |
inputs=[
|
332 |
-
|
333 |
prompt,
|
334 |
seed,
|
335 |
randomize_seed,
|
|
|
180 |
# --- UI Constants and Helpers ---
|
181 |
MAX_SEED = np.iinfo(np.int32).max
|
182 |
|
183 |
+
|
184 |
+
def concatenate_images(images, direction="horizontal"):
|
185 |
+
"""
|
186 |
+
Concatenate multiple PIL images either horizontally or vertically.
|
187 |
+
|
188 |
+
Args:
|
189 |
+
images: List of PIL Images
|
190 |
+
direction: "horizontal" or "vertical"
|
191 |
+
|
192 |
+
Returns:
|
193 |
+
PIL Image: Concatenated image
|
194 |
+
"""
|
195 |
+
if not images:
|
196 |
+
return None
|
197 |
+
|
198 |
+
# Filter out None images
|
199 |
+
valid_images = [img for img in images if img is not None]
|
200 |
+
|
201 |
+
if not valid_images:
|
202 |
+
return None
|
203 |
+
|
204 |
+
if len(valid_images) == 1:
|
205 |
+
return valid_images[0].convert("RGB")
|
206 |
+
|
207 |
+
# Convert all images to RGB
|
208 |
+
valid_images = [img.convert("RGB") for img in valid_images]
|
209 |
+
|
210 |
+
if direction == "horizontal":
|
211 |
+
# Calculate total width and max height
|
212 |
+
total_width = sum(img.width for img in valid_images)
|
213 |
+
max_height = max(img.height for img in valid_images)
|
214 |
+
|
215 |
+
# Create new image
|
216 |
+
concatenated = Image.new('RGB', (total_width, max_height), (255, 255, 255))
|
217 |
+
|
218 |
+
# Paste images
|
219 |
+
x_offset = 0
|
220 |
+
for img in valid_images:
|
221 |
+
# Center image vertically if heights differ
|
222 |
+
y_offset = (max_height - img.height) // 2
|
223 |
+
concatenated.paste(img, (x_offset, y_offset))
|
224 |
+
x_offset += img.width
|
225 |
+
|
226 |
+
else: # vertical
|
227 |
+
# Calculate max width and total height
|
228 |
+
max_width = max(img.width for img in valid_images)
|
229 |
+
total_height = sum(img.height for img in valid_images)
|
230 |
+
|
231 |
+
# Create new image
|
232 |
+
concatenated = Image.new('RGB', (max_width, total_height), (255, 255, 255))
|
233 |
+
|
234 |
+
# Paste images
|
235 |
+
y_offset = 0
|
236 |
+
for img in valid_images:
|
237 |
+
# Center image horizontally if widths differ
|
238 |
+
x_offset = (max_width - img.width) // 2
|
239 |
+
concatenated.paste(img, (x_offset, y_offset))
|
240 |
+
y_offset += img.height
|
241 |
+
|
242 |
+
return concatenated
|
243 |
+
|
244 |
# --- Main Inference Function ---
|
245 |
@spaces.GPU(duration=60)
|
246 |
def infer(
|
247 |
+
input_images,
|
248 |
prompt,
|
249 |
seed=42,
|
250 |
randomize_seed=False,
|
|
|
264 |
|
265 |
# Set up the generator for reproducibility
|
266 |
generator = torch.Generator(device=device).manual_seed(seed)
|
267 |
+
|
268 |
+
# Handle input_images - it could be a single image or a list of images
|
269 |
+
if input_images is None:
|
270 |
+
raise gr.Error("Please upload at least one image.")
|
271 |
+
|
272 |
+
# If it's a single image (not a list), convert to list
|
273 |
+
if not isinstance(input_images, list):
|
274 |
+
input_images = [input_images]
|
275 |
+
|
276 |
+
# Concatenate images horizontally
|
277 |
+
concatenated_image = concatenate_images(input_images, "horizontal")
|
278 |
+
|
279 |
+
if concatenated_image is None:
|
280 |
+
raise gr.Error("Failed to process the input images.")
|
281 |
|
282 |
print(f"Original prompt: '{prompt}'")
|
283 |
print(f"Negative Prompt: '{negative_prompt}'")
|
|
|
290 |
# Generate the edited image - always generate just 1 image
|
291 |
try:
|
292 |
images = pipe(
|
293 |
+
concatenated_image,
|
294 |
prompt=prompt,
|
295 |
negative_prompt=negative_prompt,
|
296 |
num_inference_steps=num_inference_steps,
|
|
|
343 |
|
344 |
with gr.Row():
|
345 |
with gr.Column():
|
346 |
+
input_images = gr.Gallery(
|
347 |
+
label="Upload image(s) for editing",
|
348 |
+
show_label=True,
|
349 |
+
elem_id="gallery_input",
|
350 |
+
columns=3,
|
351 |
+
rows=2,
|
352 |
+
object_fit="contain",
|
353 |
+
height="auto",
|
354 |
+
file_types=['image'],
|
355 |
+
type='pil'
|
356 |
)
|
357 |
# Changed from Gallery to Image
|
358 |
result = gr.Image(
|
|
|
410 |
triggers=[run_button.click, prompt.submit],
|
411 |
fn=infer,
|
412 |
inputs=[
|
413 |
+
input_images,
|
414 |
prompt,
|
415 |
seed,
|
416 |
randomize_seed,
|