Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -129,7 +129,8 @@ def preprocess_image(image: Image.Image,
|
|
129 |
style_name: str = "",
|
130 |
num_steps: int = 25,
|
131 |
guidance_scale: float = 5,
|
132 |
-
controlnet_conditioning_scale: float = 1.0,
|
|
|
133 |
"""
|
134 |
Preprocess the input image.
|
135 |
|
@@ -140,32 +141,35 @@ def preprocess_image(image: Image.Image,
|
|
140 |
Image.Image: The preprocessed image.
|
141 |
"""
|
142 |
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
print("image:",type(image))
|
149 |
-
|
150 |
-
prompt, negative_prompt = apply_style(style_name, prompt, negative_prompt)
|
151 |
|
152 |
-
|
153 |
-
image = pipe_control(
|
154 |
-
prompt=prompt,
|
155 |
-
negative_prompt=negative_prompt,
|
156 |
-
image=image,
|
157 |
-
num_inference_steps=num_steps,
|
158 |
-
controlnet_conditioning_scale=controlnet_conditioning_scale,
|
159 |
-
guidance_scale=guidance_scale,
|
160 |
-
width=new_width,
|
161 |
-
height=new_height).images[0]
|
162 |
|
|
|
163 |
|
164 |
-
|
165 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
166 |
|
167 |
|
168 |
-
def preprocess_images(images: List[Tuple[Image.Image, str]]) -> List[Image.Image]:
|
169 |
"""
|
170 |
Preprocess a list of input images.
|
171 |
|
@@ -177,7 +181,7 @@ def preprocess_images(images: List[Tuple[Image.Image, str]]) -> List[Image.Image
|
|
177 |
"""
|
178 |
images = [image[0] for image in images]
|
179 |
processed_images = [pipeline.preprocess_image(image) for image in images]
|
180 |
-
return processed_images
|
181 |
|
182 |
|
183 |
def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
|
@@ -327,6 +331,9 @@ def extract_glb(
|
|
327 |
return glb_path, glb_path
|
328 |
|
329 |
|
|
|
|
|
|
|
330 |
@spaces.GPU
|
331 |
def extract_gaussian(state: dict, req: gr.Request) -> Tuple[str, str]:
|
332 |
"""
|
@@ -377,11 +384,8 @@ def split_image(image: Image.Image) -> List[Image.Image]:
|
|
377 |
|
378 |
with gr.Blocks(delete_cache=(600, 600), js=js_func) as demo:
|
379 |
gr.Markdown("""
|
380 |
-
##
|
381 |
-
*
|
382 |
-
* If you find the generated 3D asset satisfactory, click "Extract GLB" to extract the GLB file and download it.
|
383 |
-
|
384 |
-
✨New: 1) Experimental multi-image support. 2) Gaussian file extraction.
|
385 |
""")
|
386 |
|
387 |
with gr.Row():
|
@@ -438,6 +442,7 @@ with gr.Blocks(delete_cache=(600, 600), js=js_func) as demo:
|
|
438 |
download_gs = gr.DownloadButton(label="Download Gaussian", interactive=False)
|
439 |
|
440 |
is_multiimage = gr.State(False)
|
|
|
441 |
output_buf = gr.State()
|
442 |
|
443 |
#Example images at the bottom of the page
|
@@ -476,15 +481,19 @@ with gr.Blocks(delete_cache=(600, 600), js=js_func) as demo:
|
|
476 |
outputs=[is_multiimage, single_image_example, multiimage_example]
|
477 |
)
|
478 |
|
479 |
-
image_prompt.upload(
|
480 |
-
|
481 |
-
|
482 |
-
|
|
|
|
|
|
|
|
|
483 |
)
|
484 |
multiimage_prompt.upload(
|
485 |
preprocess_images,
|
486 |
inputs=[multiimage_prompt],
|
487 |
-
outputs=[multiimage_prompt],
|
488 |
)
|
489 |
|
490 |
generate_btn.click(
|
@@ -493,12 +502,12 @@ with gr.Blocks(delete_cache=(600, 600), js=js_func) as demo:
|
|
493 |
outputs=[seed],
|
494 |
).then(
|
495 |
preprocess_image,
|
496 |
-
inputs=[image_prompt, prompt, negative_prompt, style],
|
497 |
-
outputs=[image_prompt],
|
498 |
).then(
|
499 |
image_to_3d,
|
500 |
-
inputs=[image_prompt, multiimage_prompt, is_multiimage, seed, ss_guidance_strength, ss_sampling_steps, slat_guidance_strength, slat_sampling_steps, multiimage_algo],
|
501 |
-
outputs=[output_buf, video_output],
|
502 |
).then(
|
503 |
lambda: tuple([gr.Button(interactive=True), gr.Button(interactive=True)]),
|
504 |
outputs=[extract_glb_btn, extract_gs_btn],
|
@@ -552,7 +561,6 @@ if __name__ == "__main__":
|
|
552 |
controlnet=controlnet,
|
553 |
vae=vae,
|
554 |
torch_dtype=torch.float16,
|
555 |
-
# scheduler=eulera_scheduler,
|
556 |
)
|
557 |
pipe_control.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe_control.scheduler.config)
|
558 |
pipe_control.to(device)
|
|
|
129 |
style_name: str = "",
|
130 |
num_steps: int = 25,
|
131 |
guidance_scale: float = 5,
|
132 |
+
controlnet_conditioning_scale: float = 1.0,
|
133 |
+
do_preprocess: bool = True) -> Image.Image:
|
134 |
"""
|
135 |
Preprocess the input image.
|
136 |
|
|
|
141 |
Image.Image: The preprocessed image.
|
142 |
"""
|
143 |
|
144 |
+
if do_preprocess:
|
145 |
+
width, height = image['composite'].size
|
146 |
+
ratio = np.sqrt(1024. * 1024. / (width * height))
|
147 |
+
new_width, new_height = int(width * ratio), int(height * ratio)
|
148 |
+
image = image['composite'].resize((new_width, new_height))
|
|
|
|
|
|
|
149 |
|
150 |
+
print("image:",type(image))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
151 |
|
152 |
+
prompt, negative_prompt = apply_style(style_name, prompt, negative_prompt)
|
153 |
|
154 |
+
print("params:", prompt, negative_prompt, style_name, num_steps, guidance_scale, controlnet_conditioning_scale)
|
155 |
+
image = pipe_control(
|
156 |
+
prompt=prompt,
|
157 |
+
negative_prompt=negative_prompt,
|
158 |
+
image=image,
|
159 |
+
num_inference_steps=num_steps,
|
160 |
+
controlnet_conditioning_scale=controlnet_conditioning_scale,
|
161 |
+
guidance_scale=guidance_scale,
|
162 |
+
width=new_width,
|
163 |
+
height=new_height).images[0]
|
164 |
+
|
165 |
+
|
166 |
+
processed_image = pipeline.preprocess_image(image)
|
167 |
+
return processed_image, False
|
168 |
+
else:
|
169 |
+
return image, False
|
170 |
|
171 |
|
172 |
+
def preprocess_images(images: List[Tuple[Image.Image, str]], do_preprocess = True) -> List[Image.Image]:
|
173 |
"""
|
174 |
Preprocess a list of input images.
|
175 |
|
|
|
181 |
"""
|
182 |
images = [image[0] for image in images]
|
183 |
processed_images = [pipeline.preprocess_image(image) for image in images]
|
184 |
+
return processed_images, False
|
185 |
|
186 |
|
187 |
def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
|
|
|
331 |
return glb_path, glb_path
|
332 |
|
333 |
|
334 |
+
def reset_do_preprocess():
|
335 |
+
return True
|
336 |
+
|
337 |
@spaces.GPU
|
338 |
def extract_gaussian(state: dict, req: gr.Request) -> Tuple[str, str]:
|
339 |
"""
|
|
|
384 |
|
385 |
with gr.Blocks(delete_cache=(600, 600), js=js_func) as demo:
|
386 |
gr.Markdown("""
|
387 |
+
## Sketch to 3D Asset with [TRELLIS](https://trellis3d.github.io/)
|
388 |
+
* draw or upload a sketch and click "Generate" to create a 3D asset.
|
|
|
|
|
|
|
389 |
""")
|
390 |
|
391 |
with gr.Row():
|
|
|
442 |
download_gs = gr.DownloadButton(label="Download Gaussian", interactive=False)
|
443 |
|
444 |
is_multiimage = gr.State(False)
|
445 |
+
do_preprocess = gr.State(True)
|
446 |
output_buf = gr.State()
|
447 |
|
448 |
#Example images at the bottom of the page
|
|
|
481 |
outputs=[is_multiimage, single_image_example, multiimage_example]
|
482 |
)
|
483 |
|
484 |
+
# image_prompt.upload(
|
485 |
+
# preprocess_image,
|
486 |
+
# inputs=[image_prompt, prompt, negative_prompt, style, do_preprocess],
|
487 |
+
# outputs=[image_prompt, do_preprocess],
|
488 |
+
# )
|
489 |
+
image_prompt.change(
|
490 |
+
reset_do_preprocess,
|
491 |
+
outputs=[do_preprocess]
|
492 |
)
|
493 |
multiimage_prompt.upload(
|
494 |
preprocess_images,
|
495 |
inputs=[multiimage_prompt],
|
496 |
+
outputs=[multiimage_prompt, do_preprocess],
|
497 |
)
|
498 |
|
499 |
generate_btn.click(
|
|
|
502 |
outputs=[seed],
|
503 |
).then(
|
504 |
preprocess_image,
|
505 |
+
inputs=[image_prompt, prompt, negative_prompt, style, do_preprocess],
|
506 |
+
outputs=[image_prompt, do_preprocess],
|
507 |
).then(
|
508 |
image_to_3d,
|
509 |
+
inputs=[image_prompt, multiimage_prompt, is_multiimage, seed, ss_guidance_strength, ss_sampling_steps, slat_guidance_strength, slat_sampling_steps, multiimage_algo, do_preprocess],
|
510 |
+
outputs=[output_buf, video_output, do_preprocess],
|
511 |
).then(
|
512 |
lambda: tuple([gr.Button(interactive=True), gr.Button(interactive=True)]),
|
513 |
outputs=[extract_glb_btn, extract_gs_btn],
|
|
|
561 |
controlnet=controlnet,
|
562 |
vae=vae,
|
563 |
torch_dtype=torch.float16,
|
|
|
564 |
)
|
565 |
pipe_control.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe_control.scheduler.config)
|
566 |
pipe_control.to(device)
|