Spaces:
Runtime error
Runtime error
pengHTYX
commited on
Commit
•
a72a0f3
1
Parent(s):
2736f7e
'update_layout'
Browse files- app.py +19 -16
- mvdiffusion/data/single_image_dataset.py +4 -4
- mvdiffusion/pipelines/pipeline_mvdiffusion_unclip.py +1 -1
- requirements.txt +1 -1
app.py
CHANGED
@@ -141,23 +141,20 @@ def preprocess(predictor, input_image, chk_group=None, segment=True, rescale=Fal
|
|
141 |
input_image = Image.fromarray((rgb * 255).astype(np.uint8))
|
142 |
else:
|
143 |
input_image = expand2square(input_image, (127, 127, 127, 0))
|
144 |
-
return input_image, input_image.resize((
|
145 |
|
146 |
|
147 |
def load_era3d_pipeline(cfg):
|
148 |
# Load scheduler, tokenizer and models.
|
149 |
|
150 |
pipeline = StableUnCLIPImg2ImgPipeline.from_pretrained(
|
151 |
-
|
152 |
-
|
153 |
)
|
154 |
|
155 |
-
# pipeline.to('cuda:0')
|
156 |
-
pipeline.unet.enable_xformers_memory_efficient_attention()
|
157 |
-
|
158 |
-
|
159 |
if torch.cuda.is_available():
|
160 |
pipeline.to('cuda:0')
|
|
|
161 |
# sys.main_lock = threading.Lock()
|
162 |
return pipeline
|
163 |
|
@@ -165,8 +162,9 @@ def load_era3d_pipeline(cfg):
|
|
165 |
from mvdiffusion.data.single_image_dataset import SingleImageDataset
|
166 |
|
167 |
|
168 |
-
def prepare_data(single_image, crop_size):
|
169 |
-
dataset = SingleImageDataset(root_dir='', num_views=6, img_wh=[512, 512], bg_color='white',
|
|
|
170 |
return dataset[0]
|
171 |
|
172 |
scene = 'scene'
|
@@ -179,7 +177,7 @@ def run_pipeline(pipeline, cfg, single_image, guidance_scale, steps, seed, crop_
|
|
179 |
if chk_group is not None:
|
180 |
write_image = "Write Results" in chk_group
|
181 |
|
182 |
-
batch = prepare_data(single_image, crop_size)
|
183 |
|
184 |
pipeline.set_progress_bar_config(disable=True)
|
185 |
seed = int(seed)
|
@@ -203,7 +201,7 @@ def run_pipeline(pipeline, cfg, single_image, guidance_scale, steps, seed, crop_
|
|
203 |
guidance_scale=guidance_scale,
|
204 |
output_type='pt',
|
205 |
num_images_per_prompt=1,
|
206 |
-
return_elevation_focal=cfg.log_elevation_focal_length,
|
207 |
**cfg.pipe_validation_kwargs
|
208 |
).images
|
209 |
|
@@ -314,6 +312,7 @@ def run_demo():
|
|
314 |
custom_css = '''#disp_image {
|
315 |
text-align: center; /* Horizontally center the content */
|
316 |
}'''
|
|
|
317 |
|
318 |
with gr.Blocks(title=_TITLE, theme=custom_theme, css=custom_css) as demo:
|
319 |
with gr.Row():
|
@@ -322,14 +321,16 @@ def run_demo():
|
|
322 |
gr.Markdown(_DESCRIPTION)
|
323 |
with gr.Row(variant='panel'):
|
324 |
with gr.Column(scale=1):
|
325 |
-
input_image = gr.Image(type='pil', image_mode='RGBA', height=
|
326 |
|
327 |
with gr.Column(scale=1):
|
|
|
|
|
328 |
processed_image = gr.Image(
|
329 |
type='pil',
|
330 |
label="Processed Image",
|
331 |
interactive=False,
|
332 |
-
height=
|
333 |
image_mode='RGBA',
|
334 |
elem_id="disp_image",
|
335 |
visible=True,
|
@@ -341,7 +342,7 @@ def run_demo():
|
|
341 |
# label="3D Model", height=320,
|
342 |
# # camera_position=[0,0,2.0]
|
343 |
# )
|
344 |
-
|
345 |
with gr.Row(variant='panel'):
|
346 |
with gr.Column(scale=1):
|
347 |
example_folder = os.path.join(os.path.dirname(__file__), "./examples")
|
@@ -391,6 +392,7 @@ def run_demo():
|
|
391 |
view_1 = gr.Image(interactive=False, height=512, show_label=False)
|
392 |
view_2 = gr.Image(interactive=False, height=512, show_label=False)
|
393 |
view_3 = gr.Image(interactive=False, height=512, show_label=False)
|
|
|
394 |
view_4 = gr.Image(interactive=False, height=512, show_label=False)
|
395 |
view_5 = gr.Image(interactive=False, height=512, show_label=False)
|
396 |
view_6 = gr.Image(interactive=False, height=512, show_label=False)
|
@@ -398,10 +400,11 @@ def run_demo():
|
|
398 |
normal_1 = gr.Image(interactive=False, height=512, show_label=False)
|
399 |
normal_2 = gr.Image(interactive=False, height=512, show_label=False)
|
400 |
normal_3 = gr.Image(interactive=False, height=512, show_label=False)
|
|
|
401 |
normal_4 = gr.Image(interactive=False, height=512, show_label=False)
|
402 |
normal_5 = gr.Image(interactive=False, height=512, show_label=False)
|
403 |
normal_6 = gr.Image(interactive=False, height=512, show_label=False)
|
404 |
-
|
405 |
run_btn.click(
|
406 |
fn=partial(preprocess, predictor), inputs=[input_image, input_processing], outputs=[processed_image_highres, processed_image], queue=True
|
407 |
).success(
|
@@ -414,7 +417,7 @@ def run_demo():
|
|
414 |
# )
|
415 |
|
416 |
demo.queue().launch(share=True, max_threads=80)
|
417 |
-
|
418 |
|
419 |
if __name__ == '__main__':
|
420 |
fire.Fire(run_demo)
|
|
|
141 |
input_image = Image.fromarray((rgb * 255).astype(np.uint8))
|
142 |
else:
|
143 |
input_image = expand2square(input_image, (127, 127, 127, 0))
|
144 |
+
return input_image, input_image.resize((320, 320), Image.Resampling.LANCZOS)
|
145 |
|
146 |
|
147 |
def load_era3d_pipeline(cfg):
|
148 |
# Load scheduler, tokenizer and models.
|
149 |
|
150 |
pipeline = StableUnCLIPImg2ImgPipeline.from_pretrained(
|
151 |
+
cfg.pretrained_model_name_or_path,
|
152 |
+
torch_dtype=weight_dtype
|
153 |
)
|
154 |
|
|
|
|
|
|
|
|
|
155 |
if torch.cuda.is_available():
|
156 |
pipeline.to('cuda:0')
|
157 |
+
pipeline.unet.enable_xformers_memory_efficient_attention()
|
158 |
# sys.main_lock = threading.Lock()
|
159 |
return pipeline
|
160 |
|
|
|
162 |
from mvdiffusion.data.single_image_dataset import SingleImageDataset
|
163 |
|
164 |
|
165 |
+
def prepare_data(single_image, crop_size, cfg):
|
166 |
+
dataset = SingleImageDataset(root_dir='', num_views=6, img_wh=[512, 512], bg_color='white',
|
167 |
+
crop_size=crop_size, single_image=single_image, prompt_embeds_path=cfg.validation_dataset.prompt_embeds_path)
|
168 |
return dataset[0]
|
169 |
|
170 |
scene = 'scene'
|
|
|
177 |
if chk_group is not None:
|
178 |
write_image = "Write Results" in chk_group
|
179 |
|
180 |
+
batch = prepare_data(single_image, crop_size, cfg)
|
181 |
|
182 |
pipeline.set_progress_bar_config(disable=True)
|
183 |
seed = int(seed)
|
|
|
201 |
guidance_scale=guidance_scale,
|
202 |
output_type='pt',
|
203 |
num_images_per_prompt=1,
|
204 |
+
# return_elevation_focal=cfg.log_elevation_focal_length,
|
205 |
**cfg.pipe_validation_kwargs
|
206 |
).images
|
207 |
|
|
|
312 |
custom_css = '''#disp_image {
|
313 |
text-align: center; /* Horizontally center the content */
|
314 |
}'''
|
315 |
+
|
316 |
|
317 |
with gr.Blocks(title=_TITLE, theme=custom_theme, css=custom_css) as demo:
|
318 |
with gr.Row():
|
|
|
321 |
gr.Markdown(_DESCRIPTION)
|
322 |
with gr.Row(variant='panel'):
|
323 |
with gr.Column(scale=1):
|
324 |
+
input_image = gr.Image(type='pil', image_mode='RGBA', height=320, label='Input image')
|
325 |
|
326 |
with gr.Column(scale=1):
|
327 |
+
processed_image_highres = gr.Image(type='pil', image_mode='RGBA', visible=False)
|
328 |
+
|
329 |
processed_image = gr.Image(
|
330 |
type='pil',
|
331 |
label="Processed Image",
|
332 |
interactive=False,
|
333 |
+
# height=320,
|
334 |
image_mode='RGBA',
|
335 |
elem_id="disp_image",
|
336 |
visible=True,
|
|
|
342 |
# label="3D Model", height=320,
|
343 |
# # camera_position=[0,0,2.0]
|
344 |
# )
|
345 |
+
|
346 |
with gr.Row(variant='panel'):
|
347 |
with gr.Column(scale=1):
|
348 |
example_folder = os.path.join(os.path.dirname(__file__), "./examples")
|
|
|
392 |
view_1 = gr.Image(interactive=False, height=512, show_label=False)
|
393 |
view_2 = gr.Image(interactive=False, height=512, show_label=False)
|
394 |
view_3 = gr.Image(interactive=False, height=512, show_label=False)
|
395 |
+
with gr.Row():
|
396 |
view_4 = gr.Image(interactive=False, height=512, show_label=False)
|
397 |
view_5 = gr.Image(interactive=False, height=512, show_label=False)
|
398 |
view_6 = gr.Image(interactive=False, height=512, show_label=False)
|
|
|
400 |
normal_1 = gr.Image(interactive=False, height=512, show_label=False)
|
401 |
normal_2 = gr.Image(interactive=False, height=512, show_label=False)
|
402 |
normal_3 = gr.Image(interactive=False, height=512, show_label=False)
|
403 |
+
with gr.Row():
|
404 |
normal_4 = gr.Image(interactive=False, height=512, show_label=False)
|
405 |
normal_5 = gr.Image(interactive=False, height=512, show_label=False)
|
406 |
normal_6 = gr.Image(interactive=False, height=512, show_label=False)
|
407 |
+
print('Launching...')
|
408 |
run_btn.click(
|
409 |
fn=partial(preprocess, predictor), inputs=[input_image, input_processing], outputs=[processed_image_highres, processed_image], queue=True
|
410 |
).success(
|
|
|
417 |
# )
|
418 |
|
419 |
demo.queue().launch(share=True, max_threads=80)
|
420 |
+
|
421 |
|
422 |
if __name__ == '__main__':
|
423 |
fire.Fire(run_demo)
|
mvdiffusion/data/single_image_dataset.py
CHANGED
@@ -236,10 +236,10 @@ class SingleImageDataset(Dataset):
|
|
236 |
color_prompt_embeddings = self.color_text_embeds if hasattr(self, 'color_text_embeds') else None
|
237 |
|
238 |
out = {
|
239 |
-
'imgs_in': img_tensors_in,
|
240 |
-
'alphas': alpha_tensors_in,
|
241 |
-
'normal_prompt_embeddings': normal_prompt_embeddings,
|
242 |
-
'color_prompt_embeddings': color_prompt_embeddings,
|
243 |
'filename': filename,
|
244 |
}
|
245 |
|
|
|
236 |
color_prompt_embeddings = self.color_text_embeds if hasattr(self, 'color_text_embeds') else None
|
237 |
|
238 |
out = {
|
239 |
+
'imgs_in': img_tensors_in.unsqueeze(0),
|
240 |
+
'alphas': alpha_tensors_in.unsqueeze(0),
|
241 |
+
'normal_prompt_embeddings': normal_prompt_embeddings.unsqueeze(0),
|
242 |
+
'color_prompt_embeddings': color_prompt_embeddings.unsqueeze(0),
|
243 |
'filename': filename,
|
244 |
}
|
245 |
|
mvdiffusion/pipelines/pipeline_mvdiffusion_unclip.py
CHANGED
@@ -239,7 +239,7 @@ class StableUnCLIPImg2ImgPipeline(DiffusionPipeline):
|
|
239 |
image_embeds = torch.cat([negative_prompt_embeds, normal_image_embeds, negative_prompt_embeds, color_image_embeds], 0)
|
240 |
|
241 |
# _____________________________vae input latents__________________________________________________
|
242 |
-
image_pt = torch.stack([TF.to_tensor(img) for img in image_pil], dim=0).to(device)
|
243 |
image_pt = image_pt * 2.0 - 1.0
|
244 |
image_latents = self.vae.encode(image_pt).latent_dist.mode() * self.vae.config.scaling_factor
|
245 |
# Note: repeat differently from official pipelines
|
|
|
239 |
image_embeds = torch.cat([negative_prompt_embeds, normal_image_embeds, negative_prompt_embeds, color_image_embeds], 0)
|
240 |
|
241 |
# _____________________________vae input latents__________________________________________________
|
242 |
+
image_pt = torch.stack([TF.to_tensor(img) for img in image_pil], dim=0).to(dtype=self.vae.dtype, device=device)
|
243 |
image_pt = image_pt * 2.0 - 1.0
|
244 |
image_latents = self.vae.encode(image_pt).latent_dist.mode() * self.vae.config.scaling_factor
|
245 |
# Note: repeat differently from official pipelines
|
requirements.txt
CHANGED
@@ -30,7 +30,7 @@ torch_efficient_distloss
|
|
30 |
tensorboard
|
31 |
rembg
|
32 |
segment_anything
|
33 |
-
gradio==
|
34 |
moviepy
|
35 |
kornia
|
36 |
fire
|
|
|
30 |
tensorboard
|
31 |
rembg
|
32 |
segment_anything
|
33 |
+
gradio==4.29.0
|
34 |
moviepy
|
35 |
kornia
|
36 |
fire
|