Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -38,10 +38,8 @@ def postprocess(output, prompt):
|
|
38 |
right = (i + 1) * slice_width if i < n - 1 else w
|
39 |
cropped_img = image.crop((left, 0, right, h))
|
40 |
|
41 |
-
# 生成 caption
|
42 |
caption = prompt[i]
|
43 |
|
44 |
-
# 存入列表
|
45 |
result.append((cropped_img, caption))
|
46 |
return result
|
47 |
|
@@ -256,9 +254,9 @@ def run_demo_server():
|
|
256 |
with gr.Row():
|
257 |
gr.Markdown('The results of semantic segmentation may be unstable because:')
|
258 |
with gr.Row():
|
259 |
-
gr.Markdown('
|
260 |
with gr.Row():
|
261 |
-
gr.Markdown('
|
262 |
with gr.Row():
|
263 |
gr.Markdown('However, we are still able to produce some high-quality semantic segmentation results, strongly demonstrating the potential of our approach.')
|
264 |
with gr.Row():
|
@@ -280,22 +278,10 @@ def run_demo_server():
|
|
280 |
undo_button = gr.Button('Undo point')
|
281 |
matting_image_reset_btn = gr.Button(value="Reset")
|
282 |
|
283 |
-
# with gr.Row():
|
284 |
-
# img_clear_button = gr.Button("Clear Cache")
|
285 |
|
286 |
with gr.Column():
|
287 |
-
# matting_image_output = gr.Image(label='Output')
|
288 |
-
# matting_image_output = gr.Image(label='Results')
|
289 |
matting_image_output = gr.Gallery(label="Results")
|
290 |
|
291 |
-
# label="Matting Output",
|
292 |
-
# type="filepath",
|
293 |
-
# show_download_button=True,
|
294 |
-
# show_share_button=True,
|
295 |
-
# interactive=False,
|
296 |
-
# elem_classes="slider",
|
297 |
-
# position=0.25,
|
298 |
-
# )
|
299 |
|
300 |
|
301 |
|
@@ -308,7 +294,6 @@ def run_demo_server():
|
|
308 |
preprocess=False,
|
309 |
queue=False,
|
310 |
).success(
|
311 |
-
# fn=process_pipe_matting,
|
312 |
fn=inf,
|
313 |
inputs=[original_image, checkbox_group, selected_points, semantic_input],
|
314 |
outputs=[matting_image_output],
|
@@ -376,13 +361,6 @@ def run_demo_server():
|
|
376 |
cache_examples=False,
|
377 |
)
|
378 |
|
379 |
-
# examples.dataset.click(
|
380 |
-
# fn=dummy
|
381 |
-
# ).success(
|
382 |
-
# fn=set_point, # Now run the actual function after inputs are populated
|
383 |
-
# inputs=[input_image, checkbox_group, selected_points_tmp, semantic_input],
|
384 |
-
# outputs=[input_image, selected_points]
|
385 |
-
# )
|
386 |
|
387 |
demo.queue(
|
388 |
api_open=False,
|
|
|
38 |
right = (i + 1) * slice_width if i < n - 1 else w
|
39 |
cropped_img = image.crop((left, 0, right, h))
|
40 |
|
|
|
41 |
caption = prompt[i]
|
42 |
|
|
|
43 |
result.append((cropped_img, caption))
|
44 |
return result
|
45 |
|
|
|
254 |
with gr.Row():
|
255 |
gr.Markdown('The results of semantic segmentation may be unstable because:')
|
256 |
with gr.Row():
|
257 |
+
gr.Markdown('- We only trained on COCO, whose quality and quantity are insufficient to meet the requirements.')
|
258 |
with gr.Row():
|
259 |
+
gr.Markdown('- Semantic segmentation is more complex than other tasks, as it requires accurately learning the relationship between semantics and objects.')
|
260 |
with gr.Row():
|
261 |
gr.Markdown('However, we are still able to produce some high-quality semantic segmentation results, strongly demonstrating the potential of our approach.')
|
262 |
with gr.Row():
|
|
|
278 |
undo_button = gr.Button('Undo point')
|
279 |
matting_image_reset_btn = gr.Button(value="Reset")
|
280 |
|
|
|
|
|
281 |
|
282 |
with gr.Column():
|
|
|
|
|
283 |
matting_image_output = gr.Gallery(label="Results")
|
284 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
285 |
|
286 |
|
287 |
|
|
|
294 |
preprocess=False,
|
295 |
queue=False,
|
296 |
).success(
|
|
|
297 |
fn=inf,
|
298 |
inputs=[original_image, checkbox_group, selected_points, semantic_input],
|
299 |
outputs=[matting_image_output],
|
|
|
361 |
cache_examples=False,
|
362 |
)
|
363 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
364 |
|
365 |
demo.queue(
|
366 |
api_open=False,
|