Spaces:
Runtime error
Runtime error
change ismultiimages logic and add file upload function
Browse files
app.py
CHANGED
|
@@ -22,55 +22,6 @@ MAX_SEED = np.iinfo(np.int32).max
|
|
| 22 |
TMP_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'tmp')
|
| 23 |
os.makedirs(TMP_DIR, exist_ok=True)
|
| 24 |
|
| 25 |
-
def to_pil_list(
|
| 26 |
-
multiimages: List[
|
| 27 |
-
Union[
|
| 28 |
-
Image.Image,
|
| 29 |
-
Tuple[Image.Image, str],
|
| 30 |
-
gr.File,
|
| 31 |
-
Tuple[gr.File, str],
|
| 32 |
-
str, # fallback: plain path
|
| 33 |
-
Path
|
| 34 |
-
]
|
| 35 |
-
]
|
| 36 |
-
) -> List[Image.Image]:
|
| 37 |
-
"""
|
| 38 |
-
Convert a heterogeneous `multiimages` list into a homogeneous
|
| 39 |
-
`List[Image.Image]`.
|
| 40 |
-
|
| 41 |
-
Accepts elements in any of the following forms:
|
| 42 |
-
• PIL.Image
|
| 43 |
-
• (PIL.Image, caption)
|
| 44 |
-
• gr.File (gr.File.name is the temp‑file path)
|
| 45 |
-
• (gr.File, caption)
|
| 46 |
-
• str / pathlib.Path (direct file path)
|
| 47 |
-
|
| 48 |
-
Returns:
|
| 49 |
-
List[Image.Image] -- guaranteed PIL images
|
| 50 |
-
"""
|
| 51 |
-
pil_imgs: List[Image.Image] = []
|
| 52 |
-
|
| 53 |
-
for item in multiimages:
|
| 54 |
-
# Unpack tuple/list, keep first element
|
| 55 |
-
if isinstance(item, (tuple, list)):
|
| 56 |
-
item = item[0]
|
| 57 |
-
|
| 58 |
-
if isinstance(item, Image.Image): # already PIL
|
| 59 |
-
pil_imgs.append(item)
|
| 60 |
-
|
| 61 |
-
elif hasattr(item, "name"): # gr.File
|
| 62 |
-
pil_imgs.append(Image.open(item.name))
|
| 63 |
-
|
| 64 |
-
elif isinstance(item, (str, Path)): # file path
|
| 65 |
-
pil_imgs.append(Image.open(item))
|
| 66 |
-
|
| 67 |
-
else:
|
| 68 |
-
raise TypeError(
|
| 69 |
-
f"Unsupported element in multiimages: {type(item)}"
|
| 70 |
-
)
|
| 71 |
-
|
| 72 |
-
return pil_imgs
|
| 73 |
-
|
| 74 |
def start_session(req: gr.Request):
|
| 75 |
user_dir = os.path.join(TMP_DIR, str(req.session_hash))
|
| 76 |
os.makedirs(user_dir, exist_ok=True)
|
|
@@ -109,6 +60,16 @@ def preprocess_images(images: List[Tuple[Image.Image, str]]) -> List[Image.Image
|
|
| 109 |
processed_images = [pipeline.preprocess_image(image) for image in images]
|
| 110 |
return processed_images
|
| 111 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 112 |
|
| 113 |
def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
|
| 114 |
return {
|
|
@@ -160,7 +121,7 @@ def get_seed(randomize_seed: bool, seed: int) -> int:
|
|
| 160 |
@spaces.GPU
|
| 161 |
def image_to_3d(
|
| 162 |
image: Image.Image,
|
| 163 |
-
multiimages: List[Tuple[Image.Image, str]],
|
| 164 |
is_multiimage: str,
|
| 165 |
seed: int,
|
| 166 |
ss_guidance_strength: float,
|
|
@@ -193,6 +154,9 @@ def image_to_3d(
|
|
| 193 |
os.makedirs(user_dir, exist_ok=True)
|
| 194 |
is_multiimage = is_multiimage.lower() == "true"
|
| 195 |
|
|
|
|
|
|
|
|
|
|
| 196 |
# Run pipeline depending on mode
|
| 197 |
if not is_multiimage:
|
| 198 |
outputs = pipeline.run(
|
|
@@ -210,7 +174,7 @@ def image_to_3d(
|
|
| 210 |
},
|
| 211 |
)
|
| 212 |
else:
|
| 213 |
-
pil_images =
|
| 214 |
outputs = pipeline.run_multi_image(
|
| 215 |
pil_images,
|
| 216 |
seed=seed,
|
|
@@ -386,8 +350,14 @@ def test_for_api_gen(image: Image.Image) -> Image.Image:
|
|
| 386 |
"""
|
| 387 |
return image
|
| 388 |
|
| 389 |
-
def update_is_multiimage(event: SelectData):
|
| 390 |
-
return "true" if event.index == 1 else "false"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 391 |
|
| 392 |
|
| 393 |
with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
@@ -428,17 +398,20 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 428 |
*NOTE: this is an experimental algorithm without training a specialized model. It may not produce the best results for all images, especially those having different poses or inconsistent details.*
|
| 429 |
""")
|
| 430 |
|
| 431 |
-
is_multiimage = gr.
|
| 432 |
-
choices=["true", "false"],
|
| 433 |
-
value="false",
|
| 434 |
-
label="Use multi-image mode",
|
| 435 |
-
visible=True
|
| 436 |
-
)
|
| 437 |
|
| 438 |
input_tabs.select(
|
| 439 |
fn=update_is_multiimage,
|
| 440 |
outputs=is_multiimage
|
| 441 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 442 |
|
| 443 |
with gr.Accordion(label="Generation Settings", open=False):
|
| 444 |
seed = gr.Slider(0, MAX_SEED, label="Seed", value=0, step=1)
|
|
@@ -466,7 +439,7 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 466 |
with gr.Row():
|
| 467 |
quick_generate_glb_btn = gr.Button("Quick Generate GLB")
|
| 468 |
quick_generate_gs_btn = gr.Button("Quick Generate Gaussian")
|
| 469 |
-
|
| 470 |
gr.Markdown("""
|
| 471 |
*NOTE: Gaussian file can be very large (~50MB), it will take a while to display and download.*
|
| 472 |
""")
|
|
@@ -499,7 +472,7 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 499 |
examples=prepare_multi_example(),
|
| 500 |
inputs=[image_prompt],
|
| 501 |
fn=split_image,
|
| 502 |
-
outputs=[
|
| 503 |
run_on_click=True,
|
| 504 |
examples_per_page=8,
|
| 505 |
)
|
|
@@ -522,12 +495,24 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 522 |
inputs=[image_prompt],
|
| 523 |
outputs=[image_prompt],
|
| 524 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 525 |
multiimage_prompt.upload(
|
| 526 |
-
preprocess_images,
|
| 527 |
inputs=[multiimage_prompt],
|
| 528 |
-
outputs=[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 529 |
)
|
| 530 |
|
|
|
|
| 531 |
generate_btn.click(
|
| 532 |
get_seed,
|
| 533 |
inputs=[randomize_seed, seed],
|
|
@@ -535,7 +520,7 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 535 |
).then(
|
| 536 |
image_to_3d,
|
| 537 |
inputs=[
|
| 538 |
-
image_prompt,
|
| 539 |
ss_guidance_strength, ss_sampling_steps,
|
| 540 |
slat_guidance_strength, slat_sampling_steps, multiimage_algo
|
| 541 |
],
|
|
@@ -577,7 +562,7 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 577 |
fn=quick_generate_glb,
|
| 578 |
inputs=[
|
| 579 |
image_prompt,
|
| 580 |
-
|
| 581 |
is_multiimage,
|
| 582 |
seed,
|
| 583 |
ss_guidance_strength,
|
|
@@ -595,7 +580,7 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 595 |
fn=quick_generate_gs,
|
| 596 |
inputs=[
|
| 597 |
image_prompt,
|
| 598 |
-
|
| 599 |
is_multiimage,
|
| 600 |
seed,
|
| 601 |
ss_guidance_strength,
|
|
@@ -606,6 +591,24 @@ with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
| 606 |
],
|
| 607 |
outputs=[model_output, download_gs],
|
| 608 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 609 |
|
| 610 |
|
| 611 |
|
|
|
|
| 22 |
TMP_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'tmp')
|
| 23 |
os.makedirs(TMP_DIR, exist_ok=True)
|
| 24 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
def start_session(req: gr.Request):
|
| 26 |
user_dir = os.path.join(TMP_DIR, str(req.session_hash))
|
| 27 |
os.makedirs(user_dir, exist_ok=True)
|
|
|
|
| 60 |
processed_images = [pipeline.preprocess_image(image) for image in images]
|
| 61 |
return processed_images
|
| 62 |
|
| 63 |
+
def preprocess_upload_images(file_list: List[Any]) -> List[Tuple[Image.Image, str]]:
|
| 64 |
+
"""
|
| 65 |
+
Resize all input images to 518x518 and return (image, filename) pairs.
|
| 66 |
+
"""
|
| 67 |
+
images = [
|
| 68 |
+
(Image.open(f.name).convert("RGBA").resize((518, 518), Image.Resampling.LANCZOS), f.name)
|
| 69 |
+
for f in file_list
|
| 70 |
+
]
|
| 71 |
+
return images
|
| 72 |
+
|
| 73 |
|
| 74 |
def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
|
| 75 |
return {
|
|
|
|
| 121 |
@spaces.GPU
|
| 122 |
def image_to_3d(
|
| 123 |
image: Image.Image,
|
| 124 |
+
multiimages: Union[List[Tuple[Image.Image, str]], List[Any]],
|
| 125 |
is_multiimage: str,
|
| 126 |
seed: int,
|
| 127 |
ss_guidance_strength: float,
|
|
|
|
| 154 |
os.makedirs(user_dir, exist_ok=True)
|
| 155 |
is_multiimage = is_multiimage.lower() == "true"
|
| 156 |
|
| 157 |
+
if multiimages and not isinstance(multiimages[0], tuple):
|
| 158 |
+
multiimages = preprocess_upload_images(multiimages)
|
| 159 |
+
|
| 160 |
# Run pipeline depending on mode
|
| 161 |
if not is_multiimage:
|
| 162 |
outputs = pipeline.run(
|
|
|
|
| 174 |
},
|
| 175 |
)
|
| 176 |
else:
|
| 177 |
+
pil_images = [d[0] for d in multiimages]
|
| 178 |
outputs = pipeline.run_multi_image(
|
| 179 |
pil_images,
|
| 180 |
seed=seed,
|
|
|
|
| 350 |
"""
|
| 351 |
return image
|
| 352 |
|
| 353 |
+
def update_is_multiimage(event: gr.SelectData):
|
| 354 |
+
return gr.update("true" if event.index == 1 else "false")
|
| 355 |
+
|
| 356 |
+
def toggle_multiimage_visibility(choice: str):
|
| 357 |
+
if choice == "true":
|
| 358 |
+
return gr.update(visible=True), gr.update(visible=False)
|
| 359 |
+
else:
|
| 360 |
+
return gr.update(visible=False), gr.update(visible=False)
|
| 361 |
|
| 362 |
|
| 363 |
with gr.Blocks(delete_cache=(600, 600)) as demo:
|
|
|
|
| 398 |
*NOTE: this is an experimental algorithm without training a specialized model. It may not produce the best results for all images, especially those having different poses or inconsistent details.*
|
| 399 |
""")
|
| 400 |
|
| 401 |
+
is_multiimage = gr.Textbox(value="false", visible=True, interactive=False, label="is_multiimage")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 402 |
|
| 403 |
input_tabs.select(
|
| 404 |
fn=update_is_multiimage,
|
| 405 |
outputs=is_multiimage
|
| 406 |
)
|
| 407 |
+
uploaded_api_images = gr.Files(file_types=["image"], label="Upload Images")
|
| 408 |
+
multiimage_combined = gr.State()
|
| 409 |
+
|
| 410 |
+
is_multiimage.change(
|
| 411 |
+
fn=toggle_multiimage_visibility,
|
| 412 |
+
inputs=is_multiimage,
|
| 413 |
+
outputs=[uploaded_api_images, multiimage_prompt]
|
| 414 |
+
)
|
| 415 |
|
| 416 |
with gr.Accordion(label="Generation Settings", open=False):
|
| 417 |
seed = gr.Slider(0, MAX_SEED, label="Seed", value=0, step=1)
|
|
|
|
| 439 |
with gr.Row():
|
| 440 |
quick_generate_glb_btn = gr.Button("Quick Generate GLB")
|
| 441 |
quick_generate_gs_btn = gr.Button("Quick Generate Gaussian")
|
| 442 |
+
|
| 443 |
gr.Markdown("""
|
| 444 |
*NOTE: Gaussian file can be very large (~50MB), it will take a while to display and download.*
|
| 445 |
""")
|
|
|
|
| 472 |
examples=prepare_multi_example(),
|
| 473 |
inputs=[image_prompt],
|
| 474 |
fn=split_image,
|
| 475 |
+
outputs=[multiimage_combined],
|
| 476 |
run_on_click=True,
|
| 477 |
examples_per_page=8,
|
| 478 |
)
|
|
|
|
| 495 |
inputs=[image_prompt],
|
| 496 |
outputs=[image_prompt],
|
| 497 |
)
|
| 498 |
+
# multiimage_prompt.upload(
|
| 499 |
+
# preprocess_images,
|
| 500 |
+
# inputs=[multiimage_prompt],
|
| 501 |
+
# outputs=[multiimage_prompt],
|
| 502 |
+
# )
|
| 503 |
multiimage_prompt.upload(
|
| 504 |
+
fn=preprocess_images,
|
| 505 |
inputs=[multiimage_prompt],
|
| 506 |
+
outputs=[multiimage_combined],
|
| 507 |
+
)
|
| 508 |
+
uploaded_api_images.upload(
|
| 509 |
+
fn=preprocess_upload_images,
|
| 510 |
+
inputs=[uploaded_api_images],
|
| 511 |
+
outputs=[multiimage_combined],
|
| 512 |
+
preprocess=False,
|
| 513 |
)
|
| 514 |
|
| 515 |
+
|
| 516 |
generate_btn.click(
|
| 517 |
get_seed,
|
| 518 |
inputs=[randomize_seed, seed],
|
|
|
|
| 520 |
).then(
|
| 521 |
image_to_3d,
|
| 522 |
inputs=[
|
| 523 |
+
image_prompt, multiimage_combined, is_multiimage, seed,
|
| 524 |
ss_guidance_strength, ss_sampling_steps,
|
| 525 |
slat_guidance_strength, slat_sampling_steps, multiimage_algo
|
| 526 |
],
|
|
|
|
| 562 |
fn=quick_generate_glb,
|
| 563 |
inputs=[
|
| 564 |
image_prompt,
|
| 565 |
+
multiimage_combined,
|
| 566 |
is_multiimage,
|
| 567 |
seed,
|
| 568 |
ss_guidance_strength,
|
|
|
|
| 580 |
fn=quick_generate_gs,
|
| 581 |
inputs=[
|
| 582 |
image_prompt,
|
| 583 |
+
multiimage_combined,
|
| 584 |
is_multiimage,
|
| 585 |
seed,
|
| 586 |
ss_guidance_strength,
|
|
|
|
| 591 |
],
|
| 592 |
outputs=[model_output, download_gs],
|
| 593 |
)
|
| 594 |
+
generate_btn.click(
|
| 595 |
+
fn=image_to_3d,
|
| 596 |
+
inputs=[
|
| 597 |
+
image_prompt, # image: Image.Image
|
| 598 |
+
multiimage_combined, # multiimages: List[UploadedFile] or List[Tuple[Image, str]]
|
| 599 |
+
is_multiimage, # is_multiimage: str
|
| 600 |
+
seed,
|
| 601 |
+
ss_guidance_strength,
|
| 602 |
+
ss_sampling_steps,
|
| 603 |
+
slat_guidance_strength,
|
| 604 |
+
slat_sampling_steps,
|
| 605 |
+
multiimage_algo,
|
| 606 |
+
],
|
| 607 |
+
outputs=[
|
| 608 |
+
output_buf,
|
| 609 |
+
video_output
|
| 610 |
+
]
|
| 611 |
+
)
|
| 612 |
|
| 613 |
|
| 614 |
|