Spaces:
Runtime error
Runtime error
update
Browse files- infer_api.py +81 -75
- infer_api_bk.py +889 -0
infer_api.py
CHANGED
|
@@ -367,13 +367,13 @@ class InferAPI:
|
|
| 367 |
continue
|
| 368 |
hf_hub_download(repo_id, file, local_dir="./ckpt")
|
| 369 |
|
| 370 |
-
self.canonical_infer = InferCanonicalAPI(self.canonical_configs)
|
| 371 |
# self.multiview_infer = InferMultiviewAPI(self.multiview_configs)
|
| 372 |
# self.slrm_infer = InferSlrmAPI(self.slrm_configs)
|
| 373 |
# self.refine_infer = InferRefineAPI(self.refine_configs)
|
| 374 |
|
| 375 |
def genStage1(self, img, seed):
|
| 376 |
-
return
|
| 377 |
|
| 378 |
def genStage2(self, img, seed, num_levels):
|
| 379 |
return self.multiview_infer.gen(img, seed, num_levels)
|
|
@@ -811,79 +811,85 @@ class InferMultiviewAPI:
|
|
| 811 |
return results
|
| 812 |
|
| 813 |
|
| 814 |
-
|
| 815 |
-
|
| 816 |
-
|
| 817 |
-
|
| 818 |
-
|
| 819 |
-
|
| 820 |
-
|
| 821 |
-
|
| 822 |
-
|
| 823 |
-
|
| 824 |
-
|
| 825 |
-
|
| 826 |
-
|
| 827 |
-
|
| 828 |
-
|
| 829 |
-
|
| 830 |
-
|
| 831 |
-
|
| 832 |
-
|
| 833 |
-
|
| 834 |
-
|
|
|
|
|
|
|
|
|
|
| 835 |
):
|
| 836 |
-
|
| 837 |
-
|
| 838 |
-
|
| 839 |
-
|
| 840 |
-
|
| 841 |
-
|
| 842 |
-
|
| 843 |
-
|
| 844 |
-
|
| 845 |
-
|
| 846 |
-
|
| 847 |
-
|
| 848 |
-
|
| 849 |
-
|
| 850 |
-
|
| 851 |
-
|
| 852 |
-
|
| 853 |
-
|
| 854 |
-
|
| 855 |
-
|
| 856 |
-
|
| 857 |
-
|
| 858 |
-
|
| 859 |
-
|
| 860 |
-
|
| 861 |
-
|
| 862 |
-
|
| 863 |
-
|
| 864 |
-
|
| 865 |
-
|
| 866 |
-
|
| 867 |
-
|
| 868 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 869 |
|
| 870 |
-
def canonicalize(self, image, seed):
|
| 871 |
-
return inference(
|
| 872 |
-
self.validation_pipeline, image, self.vae, self.feature_extractor, self.image_encoder, self.unet, self.ref_unet, self.tokenizer, self.text_encoder,
|
| 873 |
-
self.pretrained_model_path, self.validation, self.width_input, self.height_input, self.unet_condition_type,
|
| 874 |
-
use_noise=self.use_noise, noise_d=self.noise_d, crop=True, seed=seed, timestep=self.timestep
|
| 875 |
-
)
|
| 876 |
|
| 877 |
-
|
| 878 |
-
|
| 879 |
-
|
| 880 |
-
|
| 881 |
-
|
| 882 |
-
|
| 883 |
-
|
| 884 |
-
|
| 885 |
-
|
| 886 |
-
|
| 887 |
-
|
| 888 |
-
|
| 889 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 367 |
continue
|
| 368 |
hf_hub_download(repo_id, file, local_dir="./ckpt")
|
| 369 |
|
| 370 |
+
# self.canonical_infer = InferCanonicalAPI(self.canonical_configs)
|
| 371 |
# self.multiview_infer = InferMultiviewAPI(self.multiview_configs)
|
| 372 |
# self.slrm_infer = InferSlrmAPI(self.slrm_configs)
|
| 373 |
# self.refine_infer = InferRefineAPI(self.refine_configs)
|
| 374 |
|
| 375 |
def genStage1(self, img, seed):
|
| 376 |
+
return infer_canonicalize_gen(img, seed)
|
| 377 |
|
| 378 |
def genStage2(self, img, seed, num_levels):
|
| 379 |
return self.multiview_infer.gen(img, seed, num_levels)
|
|
|
|
| 811 |
return results
|
| 812 |
|
| 813 |
|
| 814 |
+
infer_canonicalize_config = {
|
| 815 |
+
'config_path': './configs/canonicalization-infer.yaml',
|
| 816 |
+
}
|
| 817 |
+
infer_canonicalize_device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
| 818 |
+
# print device stderr
|
| 819 |
+
import sys
|
| 820 |
+
print(f"Using device!!!!!!!!!!!!: {infer_canonicalize_device}", file=sys.stderr)
|
| 821 |
+
|
| 822 |
+
infer_canonicalize_config_path = infer_canonicalize_config['config_path']
|
| 823 |
+
infer_canonicalize_loaded_config = OmegaConf.load(infer_canonicalize_config_path)
|
| 824 |
+
|
| 825 |
+
# infer_canonicalize_setup(**infer_canonicalize_loaded_config)
|
| 826 |
+
|
| 827 |
+
def infer_canonicalize_setup(
|
| 828 |
+
validation: Dict,
|
| 829 |
+
pretrained_model_path: str,
|
| 830 |
+
local_crossattn: bool = True,
|
| 831 |
+
unet_from_pretrained_kwargs=None,
|
| 832 |
+
unet_condition_type=None,
|
| 833 |
+
use_noise=True,
|
| 834 |
+
noise_d=256,
|
| 835 |
+
timestep: int = 40,
|
| 836 |
+
width_input: int = 640,
|
| 837 |
+
height_input: int = 1024,
|
| 838 |
):
|
| 839 |
+
infer_canonicalize_width_input = width_input
|
| 840 |
+
infer_canonicalize_height_input = height_input
|
| 841 |
+
infer_canonicalize_timestep = timestep
|
| 842 |
+
infer_canonicalize_use_noise = use_noise
|
| 843 |
+
infer_canonicalize_noise_d = noise_d
|
| 844 |
+
infer_canonicalize_validation = validation
|
| 845 |
+
infer_canonicalize_unet_condition_type = unet_condition_type
|
| 846 |
+
infer_canonicalize_pretrained_model_path = pretrained_model_path
|
| 847 |
+
infer_canonicalize_local_crossattn = local_crossattn
|
| 848 |
+
infer_canonicalize_unet_from_pretrained_kwargs = unet_from_pretrained_kwargs
|
| 849 |
+
return infer_canonicalize_width_input, infer_canonicalize_height_input, infer_canonicalize_timestep, infer_canonicalize_use_noise, infer_canonicalize_noise_d, infer_canonicalize_validation, infer_canonicalize_unet_condition_type, infer_canonicalize_pretrained_model_path, infer_canonicalize_local_crossattn, infer_canonicalize_unet_from_pretrained_kwargs
|
| 850 |
+
|
| 851 |
+
infer_canonicalize_width_input, infer_canonicalize_height_input, infer_canonicalize_timestep, infer_canonicalize_use_noise, infer_canonicalize_noise_d, infer_canonicalize_validation, infer_canonicalize_unet_condition_type, infer_canonicalize_pretrained_model_path, infer_canonicalize_local_crossattn, infer_canonicalize_unet_from_pretrained_kwargs = infer_canonicalize_setup(**infer_canonicalize_loaded_config)
|
| 852 |
+
|
| 853 |
+
infer_canonicalize_tokenizer = CLIPTokenizer.from_pretrained(infer_canonicalize_pretrained_model_path, subfolder="tokenizer")
|
| 854 |
+
infer_canonicalize_text_encoder = CLIPTextModel.from_pretrained(infer_canonicalize_pretrained_model_path, subfolder="text_encoder")
|
| 855 |
+
infer_canonicalize_image_encoder = CLIPVisionModelWithProjection.from_pretrained(infer_canonicalize_pretrained_model_path, subfolder="image_encoder")
|
| 856 |
+
infer_canonicalize_feature_extractor = CLIPImageProcessor()
|
| 857 |
+
infer_canonicalize_vae = AutoencoderKL.from_pretrained(infer_canonicalize_pretrained_model_path, subfolder="vae")
|
| 858 |
+
infer_canonicalize_unet = UNetMV2DConditionModel.from_pretrained_2d(infer_canonicalize_pretrained_model_path, subfolder="unet", local_crossattn=infer_canonicalize_local_crossattn, **infer_canonicalize_unet_from_pretrained_kwargs)
|
| 859 |
+
infer_canonicalize_ref_unet = UNetMV2DRefModel.from_pretrained_2d(infer_canonicalize_pretrained_model_path, subfolder="ref_unet", local_crossattn=infer_canonicalize_local_crossattn, **infer_canonicalize_unet_from_pretrained_kwargs)
|
| 860 |
+
|
| 861 |
+
infer_canonicalize_text_encoder.to(device, dtype=weight_dtype)
|
| 862 |
+
infer_canonicalize_image_encoder.to(device, dtype=weight_dtype)
|
| 863 |
+
infer_canonicalize_vae.to(device, dtype=weight_dtype)
|
| 864 |
+
infer_canonicalize_ref_unet.to(device, dtype=weight_dtype)
|
| 865 |
+
infer_canonicalize_unet.to(device, dtype=weight_dtype)
|
| 866 |
+
|
| 867 |
+
infer_canonicalize_vae.requires_grad_(False)
|
| 868 |
+
infer_canonicalize_ref_unet.requires_grad_(False)
|
| 869 |
+
infer_canonicalize_unet.requires_grad_(False)
|
| 870 |
+
|
| 871 |
+
infer_canonicalize_noise_scheduler = DDIMScheduler.from_pretrained(infer_canonicalize_pretrained_model_path, subfolder="scheduler-zerosnr")
|
| 872 |
+
infer_canonicalize_validation_pipeline = CanonicalizationPipeline(
|
| 873 |
+
vae=infer_canonicalize_vae, text_encoder=infer_canonicalize_text_encoder, tokenizer=infer_canonicalize_tokenizer, unet=infer_canonicalize_unet, ref_unet=infer_canonicalize_ref_unet,feature_extractor=infer_canonicalize_feature_extractor,image_encoder=infer_canonicalize_image_encoder,
|
| 874 |
+
scheduler=infer_canonicalize_noise_scheduler
|
| 875 |
+
)
|
| 876 |
+
infer_canonicalize_validation_pipeline.set_progress_bar_config(disable=True)
|
| 877 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 878 |
|
| 879 |
+
def infer_canonicalize_gen(img_input, seed=0):
|
| 880 |
+
if np.array(img_input).shape[-1] == 4 and np.array(img_input)[..., 3].min() == 255:
|
| 881 |
+
# convert to RGB
|
| 882 |
+
img_input = img_input.convert("RGB")
|
| 883 |
+
img_output = inference(
|
| 884 |
+
infer_canonicalize_validation_pipeline, img_input, infer_canonicalize_vae, infer_canonicalize_feature_extractor, infer_canonicalize_image_encoder, infer_canonicalize_unet, infer_canonicalize_ref_unet, infer_canonicalize_tokenizer, infer_canonicalize_text_encoder,
|
| 885 |
+
infer_canonicalize_pretrained_model_path, infer_canonicalize_validation, infer_canonicalize_width_input, infer_canonicalize_height_input, infer_canonicalize_unet_condition_type,
|
| 886 |
+
use_noise=infer_canonicalize_use_noise, noise_d=infer_canonicalize_noise_d, crop=True, seed=seed, timestep=infer_canonicalize_timestep
|
| 887 |
+
)
|
| 888 |
+
|
| 889 |
+
max_dim = max(img_output.width, img_output.height)
|
| 890 |
+
new_image = Image.new("RGBA", (max_dim, max_dim))
|
| 891 |
+
left = (max_dim - img_output.width) // 2
|
| 892 |
+
top = (max_dim - img_output.height) // 2
|
| 893 |
+
new_image.paste(img_output, (left, top))
|
| 894 |
+
|
| 895 |
+
return new_image
|
infer_api_bk.py
ADDED
|
@@ -0,0 +1,889 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import spaces
|
| 2 |
+
from PIL import Image
|
| 3 |
+
|
| 4 |
+
import io
|
| 5 |
+
import argparse
|
| 6 |
+
import os
|
| 7 |
+
import random
|
| 8 |
+
import tempfile
|
| 9 |
+
from typing import Dict, Optional, Tuple
|
| 10 |
+
from omegaconf import OmegaConf
|
| 11 |
+
import numpy as np
|
| 12 |
+
|
| 13 |
+
import torch
|
| 14 |
+
|
| 15 |
+
from diffusers import AutoencoderKL, DDIMScheduler
|
| 16 |
+
from diffusers.utils import check_min_version
|
| 17 |
+
from tqdm.auto import tqdm
|
| 18 |
+
from transformers import CLIPTextModel, CLIPTokenizer, CLIPImageProcessor, CLIPVisionModelWithProjection
|
| 19 |
+
from torchvision import transforms
|
| 20 |
+
|
| 21 |
+
from canonicalize.models.unet_mv2d_condition import UNetMV2DConditionModel
|
| 22 |
+
from canonicalize.models.unet_mv2d_ref import UNetMV2DRefModel
|
| 23 |
+
from canonicalize.pipeline_canonicalize import CanonicalizationPipeline
|
| 24 |
+
from einops import rearrange
|
| 25 |
+
from torchvision.utils import save_image
|
| 26 |
+
import json
|
| 27 |
+
import cv2
|
| 28 |
+
|
| 29 |
+
import onnxruntime as rt
|
| 30 |
+
from huggingface_hub.file_download import hf_hub_download
|
| 31 |
+
from huggingface_hub import list_repo_files
|
| 32 |
+
from rm_anime_bg.cli import get_mask, SCALE
|
| 33 |
+
|
| 34 |
+
import argparse
|
| 35 |
+
import os
|
| 36 |
+
import cv2
|
| 37 |
+
import glob
|
| 38 |
+
import numpy as np
|
| 39 |
+
import matplotlib.pyplot as plt
|
| 40 |
+
from typing import Dict, Optional, List
|
| 41 |
+
from omegaconf import OmegaConf, DictConfig
|
| 42 |
+
from PIL import Image
|
| 43 |
+
from pathlib import Path
|
| 44 |
+
from dataclasses import dataclass
|
| 45 |
+
from typing import Dict
|
| 46 |
+
import torch
|
| 47 |
+
import torch.nn.functional as F
|
| 48 |
+
import torch.utils.checkpoint
|
| 49 |
+
import torchvision.transforms.functional as TF
|
| 50 |
+
from torch.utils.data import Dataset, DataLoader
|
| 51 |
+
from torchvision import transforms
|
| 52 |
+
from torchvision.utils import make_grid, save_image
|
| 53 |
+
from accelerate.utils import set_seed
|
| 54 |
+
from tqdm.auto import tqdm
|
| 55 |
+
from einops import rearrange, repeat
|
| 56 |
+
from multiview.pipeline_multiclass import StableUnCLIPImg2ImgPipeline
|
| 57 |
+
|
| 58 |
+
import os
|
| 59 |
+
import imageio
|
| 60 |
+
import numpy as np
|
| 61 |
+
import torch
|
| 62 |
+
import cv2
|
| 63 |
+
import glob
|
| 64 |
+
import matplotlib.pyplot as plt
|
| 65 |
+
from PIL import Image
|
| 66 |
+
from torchvision.transforms import v2
|
| 67 |
+
from pytorch_lightning import seed_everything
|
| 68 |
+
from omegaconf import OmegaConf
|
| 69 |
+
from tqdm import tqdm
|
| 70 |
+
|
| 71 |
+
from slrm.utils.train_util import instantiate_from_config
|
| 72 |
+
from slrm.utils.camera_util import (
|
| 73 |
+
FOV_to_intrinsics,
|
| 74 |
+
get_circular_camera_poses,
|
| 75 |
+
)
|
| 76 |
+
from slrm.utils.mesh_util import save_obj, save_glb
|
| 77 |
+
from slrm.utils.infer_util import images_to_video
|
| 78 |
+
|
| 79 |
+
import cv2
|
| 80 |
+
import numpy as np
|
| 81 |
+
import os
|
| 82 |
+
import trimesh
|
| 83 |
+
import argparse
|
| 84 |
+
import torch
|
| 85 |
+
import scipy
|
| 86 |
+
from PIL import Image
|
| 87 |
+
|
| 88 |
+
from refine.mesh_refine import geo_refine
|
| 89 |
+
from refine.func import make_star_cameras_orthographic
|
| 90 |
+
from refine.render import NormalsRenderer, calc_vertex_normals
|
| 91 |
+
|
| 92 |
+
import pytorch3d
|
| 93 |
+
from pytorch3d.structures import Meshes
|
| 94 |
+
from sklearn.neighbors import KDTree
|
| 95 |
+
|
| 96 |
+
from segment_anything import SamAutomaticMaskGenerator, sam_model_registry
|
| 97 |
+
|
| 98 |
+
check_min_version("0.24.0")
|
| 99 |
+
weight_dtype = torch.float16
|
| 100 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
| 101 |
+
VIEWS = ['front', 'front_right', 'right', 'back', 'left', 'front_left']
|
| 102 |
+
|
| 103 |
+
|
| 104 |
+
@spaces.GPU
|
| 105 |
+
def set_seed(seed):
|
| 106 |
+
random.seed(seed)
|
| 107 |
+
np.random.seed(seed)
|
| 108 |
+
torch.manual_seed(seed)
|
| 109 |
+
torch.cuda.manual_seed_all(seed)
|
| 110 |
+
|
| 111 |
+
|
| 112 |
+
session_infer_path = hf_hub_download(
|
| 113 |
+
repo_id="skytnt/anime-seg", filename="isnetis.onnx",
|
| 114 |
+
)
|
| 115 |
+
providers: list[str] = ["CPUExecutionProvider"]
|
| 116 |
+
if "CUDAExecutionProvider" in rt.get_available_providers():
|
| 117 |
+
providers = ["CUDAExecutionProvider"]
|
| 118 |
+
|
| 119 |
+
bkg_remover_session_infer = rt.InferenceSession(
|
| 120 |
+
session_infer_path, providers=providers,
|
| 121 |
+
)
|
| 122 |
+
|
| 123 |
+
@spaces.GPU
|
| 124 |
+
def remove_background(
|
| 125 |
+
img: np.ndarray,
|
| 126 |
+
alpha_min: float,
|
| 127 |
+
alpha_max: float,
|
| 128 |
+
) -> list:
|
| 129 |
+
img = np.array(img)
|
| 130 |
+
mask = get_mask(bkg_remover_session_infer, img)
|
| 131 |
+
mask[mask < alpha_min] = 0.0
|
| 132 |
+
mask[mask > alpha_max] = 1.0
|
| 133 |
+
img_after = (mask * img).astype(np.uint8)
|
| 134 |
+
mask = (mask * SCALE).astype(np.uint8)
|
| 135 |
+
img_after = np.concatenate([img_after, mask], axis=2, dtype=np.uint8)
|
| 136 |
+
return Image.fromarray(img_after)
|
| 137 |
+
|
| 138 |
+
|
| 139 |
+
def process_image(image, totensor, width, height):
|
| 140 |
+
assert image.mode == "RGBA"
|
| 141 |
+
|
| 142 |
+
# Find non-transparent pixels
|
| 143 |
+
non_transparent = np.nonzero(np.array(image)[..., 3])
|
| 144 |
+
min_x, max_x = non_transparent[1].min(), non_transparent[1].max()
|
| 145 |
+
min_y, max_y = non_transparent[0].min(), non_transparent[0].max()
|
| 146 |
+
image = image.crop((min_x, min_y, max_x, max_y))
|
| 147 |
+
|
| 148 |
+
# paste to center
|
| 149 |
+
max_dim = max(image.width, image.height)
|
| 150 |
+
max_height = int(max_dim * 1.2)
|
| 151 |
+
max_width = int(max_dim / (height/width) * 1.2)
|
| 152 |
+
new_image = Image.new("RGBA", (max_width, max_height))
|
| 153 |
+
left = (max_width - image.width) // 2
|
| 154 |
+
top = (max_height - image.height) // 2
|
| 155 |
+
new_image.paste(image, (left, top))
|
| 156 |
+
|
| 157 |
+
image = new_image.resize((width, height), resample=Image.BICUBIC)
|
| 158 |
+
image = np.array(image)
|
| 159 |
+
image = image.astype(np.float32) / 255.
|
| 160 |
+
assert image.shape[-1] == 4 # RGBA
|
| 161 |
+
alpha = image[..., 3:4]
|
| 162 |
+
bg_color = np.array([1., 1., 1.], dtype=np.float32)
|
| 163 |
+
image = image[..., :3] * alpha + bg_color * (1 - alpha)
|
| 164 |
+
return totensor(image)
|
| 165 |
+
|
| 166 |
+
|
| 167 |
+
@spaces.GPU
|
| 168 |
+
@torch.no_grad()
|
| 169 |
+
def inference(validation_pipeline, input_image, vae, feature_extractor, image_encoder, unet, ref_unet, tokenizer,
|
| 170 |
+
text_encoder, pretrained_model_path, validation, val_width, val_height, unet_condition_type,
|
| 171 |
+
use_noise=True, noise_d=256, crop=False, seed=100, timestep=20):
|
| 172 |
+
set_seed(seed)
|
| 173 |
+
generator = torch.Generator(device=device).manual_seed(seed)
|
| 174 |
+
|
| 175 |
+
totensor = transforms.ToTensor()
|
| 176 |
+
|
| 177 |
+
prompts = "high quality, best quality"
|
| 178 |
+
prompt_ids = tokenizer(
|
| 179 |
+
prompts, max_length=tokenizer.model_max_length, padding="max_length", truncation=True,
|
| 180 |
+
return_tensors="pt"
|
| 181 |
+
).input_ids[0]
|
| 182 |
+
|
| 183 |
+
# (B*Nv, 3, H, W)
|
| 184 |
+
B = 1
|
| 185 |
+
if input_image.mode != "RGBA":
|
| 186 |
+
# remove background
|
| 187 |
+
input_image = remove_background(input_image, 0.1, 0.9)
|
| 188 |
+
imgs_in = process_image(input_image, totensor, val_width, val_height)
|
| 189 |
+
imgs_in = rearrange(imgs_in.unsqueeze(0).unsqueeze(0), "B Nv C H W -> (B Nv) C H W")
|
| 190 |
+
|
| 191 |
+
with torch.autocast('cuda' if torch.cuda.is_available() else 'cpu', dtype=weight_dtype):
|
| 192 |
+
imgs_in = imgs_in.to(device=device)
|
| 193 |
+
# B*Nv images
|
| 194 |
+
out = validation_pipeline(prompt=prompts, image=imgs_in.to(weight_dtype), generator=generator,
|
| 195 |
+
num_inference_steps=timestep, prompt_ids=prompt_ids,
|
| 196 |
+
height=val_height, width=val_width, unet_condition_type=unet_condition_type,
|
| 197 |
+
use_noise=use_noise, **validation,)
|
| 198 |
+
out = rearrange(out, "B C f H W -> (B f) C H W", f=1)
|
| 199 |
+
|
| 200 |
+
print("OUT!!!!!!")
|
| 201 |
+
|
| 202 |
+
img_buf = io.BytesIO()
|
| 203 |
+
save_image(out[0], img_buf, format='PNG')
|
| 204 |
+
img_buf.seek(0)
|
| 205 |
+
img = Image.open(img_buf)
|
| 206 |
+
|
| 207 |
+
print("OUT2!!!!!!")
|
| 208 |
+
|
| 209 |
+
torch.cuda.empty_cache()
|
| 210 |
+
return img
|
| 211 |
+
|
| 212 |
+
|
| 213 |
+
######### Multi View Part #############
|
| 214 |
+
weight_dtype = torch.float16
|
| 215 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
| 216 |
+
|
| 217 |
+
def tensor_to_numpy(tensor):
|
| 218 |
+
return tensor.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to("cpu", torch.uint8).numpy()
|
| 219 |
+
|
| 220 |
+
|
| 221 |
+
@dataclass
|
| 222 |
+
class TestConfig:
|
| 223 |
+
pretrained_model_name_or_path: str
|
| 224 |
+
pretrained_unet_path:Optional[str]
|
| 225 |
+
revision: Optional[str]
|
| 226 |
+
validation_dataset: Dict
|
| 227 |
+
save_dir: str
|
| 228 |
+
seed: Optional[int]
|
| 229 |
+
validation_batch_size: int
|
| 230 |
+
dataloader_num_workers: int
|
| 231 |
+
save_mode: str
|
| 232 |
+
local_rank: int
|
| 233 |
+
|
| 234 |
+
pipe_kwargs: Dict
|
| 235 |
+
pipe_validation_kwargs: Dict
|
| 236 |
+
unet_from_pretrained_kwargs: Dict
|
| 237 |
+
validation_grid_nrow: int
|
| 238 |
+
camera_embedding_lr_mult: float
|
| 239 |
+
|
| 240 |
+
num_views: int
|
| 241 |
+
camera_embedding_type: str
|
| 242 |
+
|
| 243 |
+
pred_type: str
|
| 244 |
+
regress_elevation: bool
|
| 245 |
+
enable_xformers_memory_efficient_attention: bool
|
| 246 |
+
|
| 247 |
+
cond_on_normals: bool
|
| 248 |
+
cond_on_colors: bool
|
| 249 |
+
|
| 250 |
+
regress_elevation: bool
|
| 251 |
+
regress_focal_length: bool
|
| 252 |
+
|
| 253 |
+
|
| 254 |
+
|
| 255 |
+
def convert_to_numpy(tensor):
|
| 256 |
+
return tensor.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to("cpu", torch.uint8).numpy()
|
| 257 |
+
|
| 258 |
+
def save_image(tensor):
|
| 259 |
+
ndarr = convert_to_numpy(tensor)
|
| 260 |
+
return save_image_numpy(ndarr)
|
| 261 |
+
|
| 262 |
+
def save_image_numpy(ndarr):
|
| 263 |
+
im = Image.fromarray(ndarr)
|
| 264 |
+
# pad to square
|
| 265 |
+
if im.size[0] != im.size[1]:
|
| 266 |
+
size = max(im.size)
|
| 267 |
+
new_im = Image.new("RGB", (size, size))
|
| 268 |
+
# set to white
|
| 269 |
+
new_im.paste((255, 255, 255), (0, 0, size, size))
|
| 270 |
+
new_im.paste(im, ((size - im.size[0]) // 2, (size - im.size[1]) // 2))
|
| 271 |
+
im = new_im
|
| 272 |
+
# resize to 1024x1024
|
| 273 |
+
im = im.resize((1024, 1024), Image.LANCZOS)
|
| 274 |
+
return im
|
| 275 |
+
|
| 276 |
+
@spaces.GPU
|
| 277 |
+
def run_multiview_infer(data, pipeline, cfg: TestConfig, num_levels=3):
|
| 278 |
+
if cfg.seed is None:
|
| 279 |
+
generator = None
|
| 280 |
+
else:
|
| 281 |
+
generator = torch.Generator(device=pipeline.unet.device).manual_seed(cfg.seed)
|
| 282 |
+
|
| 283 |
+
images_cond = []
|
| 284 |
+
results = {}
|
| 285 |
+
|
| 286 |
+
torch.cuda.empty_cache()
|
| 287 |
+
images_cond.append(data['image_cond_rgb'][:, 0].cuda())
|
| 288 |
+
imgs_in = torch.cat([data['image_cond_rgb']]*2, dim=0).cuda()
|
| 289 |
+
num_views = imgs_in.shape[1]
|
| 290 |
+
imgs_in = rearrange(imgs_in, "B Nv C H W -> (B Nv) C H W")# (B*Nv, 3, H, W)
|
| 291 |
+
|
| 292 |
+
target_h, target_w = imgs_in.shape[-2], imgs_in.shape[-1]
|
| 293 |
+
|
| 294 |
+
normal_prompt_embeddings, clr_prompt_embeddings = data['normal_prompt_embeddings'].cuda(), data['color_prompt_embeddings'].cuda()
|
| 295 |
+
prompt_embeddings = torch.cat([normal_prompt_embeddings, clr_prompt_embeddings], dim=0)
|
| 296 |
+
prompt_embeddings = rearrange(prompt_embeddings, "B Nv N C -> (B Nv) N C")
|
| 297 |
+
|
| 298 |
+
# B*Nv images
|
| 299 |
+
unet_out = pipeline(
|
| 300 |
+
imgs_in, None, prompt_embeds=prompt_embeddings,
|
| 301 |
+
generator=generator, guidance_scale=3.0, output_type='pt', num_images_per_prompt=1,
|
| 302 |
+
height=cfg.height, width=cfg.width,
|
| 303 |
+
num_inference_steps=40, eta=1.0,
|
| 304 |
+
num_levels=num_levels,
|
| 305 |
+
)
|
| 306 |
+
|
| 307 |
+
for level in range(num_levels):
|
| 308 |
+
out = unet_out[level].images
|
| 309 |
+
bsz = out.shape[0] // 2
|
| 310 |
+
|
| 311 |
+
normals_pred = out[:bsz]
|
| 312 |
+
images_pred = out[bsz:]
|
| 313 |
+
|
| 314 |
+
if num_levels == 2:
|
| 315 |
+
results[level+1] = {'normals': [], 'images': []}
|
| 316 |
+
else:
|
| 317 |
+
results[level] = {'normals': [], 'images': []}
|
| 318 |
+
|
| 319 |
+
for i in range(bsz//num_views):
|
| 320 |
+
img_in_ = images_cond[-1][i].to(out.device)
|
| 321 |
+
for j in range(num_views):
|
| 322 |
+
view = VIEWS[j]
|
| 323 |
+
idx = i*num_views + j
|
| 324 |
+
normal = normals_pred[idx]
|
| 325 |
+
color = images_pred[idx]
|
| 326 |
+
|
| 327 |
+
## save color and normal---------------------
|
| 328 |
+
new_normal = save_image(normal)
|
| 329 |
+
new_color = save_image(color)
|
| 330 |
+
|
| 331 |
+
if num_levels == 2:
|
| 332 |
+
results[level+1]['normals'].append(new_normal)
|
| 333 |
+
results[level+1]['images'].append(new_color)
|
| 334 |
+
else:
|
| 335 |
+
results[level]['normals'].append(new_normal)
|
| 336 |
+
results[level]['images'].append(new_color)
|
| 337 |
+
|
| 338 |
+
torch.cuda.empty_cache()
|
| 339 |
+
return results
|
| 340 |
+
|
| 341 |
+
@spaces.GPU
|
| 342 |
+
def load_multiview_pipeline(cfg):
|
| 343 |
+
pipeline = StableUnCLIPImg2ImgPipeline.from_pretrained(
|
| 344 |
+
cfg.pretrained_path,
|
| 345 |
+
torch_dtype=torch.float16,)
|
| 346 |
+
pipeline.unet.enable_xformers_memory_efficient_attention()
|
| 347 |
+
if torch.cuda.is_available():
|
| 348 |
+
pipeline.to(device)
|
| 349 |
+
return pipeline
|
| 350 |
+
|
| 351 |
+
|
| 352 |
+
class InferAPI:
|
| 353 |
+
def __init__(self,
|
| 354 |
+
canonical_configs,
|
| 355 |
+
multiview_configs,
|
| 356 |
+
slrm_configs,
|
| 357 |
+
refine_configs):
|
| 358 |
+
self.canonical_configs = canonical_configs
|
| 359 |
+
self.multiview_configs = multiview_configs
|
| 360 |
+
self.slrm_configs = slrm_configs
|
| 361 |
+
self.refine_configs = refine_configs
|
| 362 |
+
|
| 363 |
+
repo_id = "hyz317/StdGEN"
|
| 364 |
+
all_files = list_repo_files(repo_id, revision="main")
|
| 365 |
+
for file in all_files:
|
| 366 |
+
if os.path.exists(file):
|
| 367 |
+
continue
|
| 368 |
+
hf_hub_download(repo_id, file, local_dir="./ckpt")
|
| 369 |
+
|
| 370 |
+
self.canonical_infer = InferCanonicalAPI(self.canonical_configs)
|
| 371 |
+
# self.multiview_infer = InferMultiviewAPI(self.multiview_configs)
|
| 372 |
+
# self.slrm_infer = InferSlrmAPI(self.slrm_configs)
|
| 373 |
+
# self.refine_infer = InferRefineAPI(self.refine_configs)
|
| 374 |
+
|
| 375 |
+
def genStage1(self, img, seed):
|
| 376 |
+
return self.canonical_infer.gen(img, seed)
|
| 377 |
+
|
| 378 |
+
def genStage2(self, img, seed, num_levels):
|
| 379 |
+
return self.multiview_infer.gen(img, seed, num_levels)
|
| 380 |
+
|
| 381 |
+
def genStage3(self, img):
|
| 382 |
+
return self.slrm_infer.gen(img)
|
| 383 |
+
|
| 384 |
+
def genStage4(self, meshes, imgs):
|
| 385 |
+
return self.refine_infer.refine(meshes, imgs)
|
| 386 |
+
|
| 387 |
+
|
| 388 |
+
############## Refine ##############
|
| 389 |
+
def fix_vert_color_glb(mesh_path):
|
| 390 |
+
from pygltflib import GLTF2, Material, PbrMetallicRoughness
|
| 391 |
+
obj1 = GLTF2().load(mesh_path)
|
| 392 |
+
obj1.meshes[0].primitives[0].material = 0
|
| 393 |
+
obj1.materials.append(Material(
|
| 394 |
+
pbrMetallicRoughness = PbrMetallicRoughness(
|
| 395 |
+
baseColorFactor = [1.0, 1.0, 1.0, 1.0],
|
| 396 |
+
metallicFactor = 0.,
|
| 397 |
+
roughnessFactor = 1.0,
|
| 398 |
+
),
|
| 399 |
+
emissiveFactor = [0.0, 0.0, 0.0],
|
| 400 |
+
doubleSided = True,
|
| 401 |
+
))
|
| 402 |
+
obj1.save(mesh_path)
|
| 403 |
+
|
| 404 |
+
|
| 405 |
+
def srgb_to_linear(c_srgb):
|
| 406 |
+
c_linear = np.where(c_srgb <= 0.04045, c_srgb / 12.92, ((c_srgb + 0.055) / 1.055) ** 2.4)
|
| 407 |
+
return c_linear.clip(0, 1.)
|
| 408 |
+
|
| 409 |
+
|
| 410 |
+
def save_py3dmesh_with_trimesh_fast(meshes: Meshes, save_glb_path, apply_sRGB_to_LinearRGB=True):
|
| 411 |
+
# convert from pytorch3d meshes to trimesh mesh
|
| 412 |
+
vertices = meshes.verts_packed().cpu().float().numpy()
|
| 413 |
+
triangles = meshes.faces_packed().cpu().long().numpy()
|
| 414 |
+
np_color = meshes.textures.verts_features_packed().cpu().float().numpy()
|
| 415 |
+
if save_glb_path.endswith(".glb"):
|
| 416 |
+
# rotate 180 along +Y
|
| 417 |
+
vertices[:, [0, 2]] = -vertices[:, [0, 2]]
|
| 418 |
+
|
| 419 |
+
if apply_sRGB_to_LinearRGB:
|
| 420 |
+
np_color = srgb_to_linear(np_color)
|
| 421 |
+
assert vertices.shape[0] == np_color.shape[0]
|
| 422 |
+
assert np_color.shape[1] == 3
|
| 423 |
+
assert 0 <= np_color.min() and np_color.max() <= 1.001, f"min={np_color.min()}, max={np_color.max()}"
|
| 424 |
+
np_color = np.clip(np_color, 0, 1)
|
| 425 |
+
mesh = trimesh.Trimesh(vertices=vertices, faces=triangles, vertex_colors=np_color)
|
| 426 |
+
mesh.remove_unreferenced_vertices()
|
| 427 |
+
# save mesh
|
| 428 |
+
mesh.export(save_glb_path)
|
| 429 |
+
if save_glb_path.endswith(".glb"):
|
| 430 |
+
fix_vert_color_glb(save_glb_path)
|
| 431 |
+
print(f"saving to {save_glb_path}")
|
| 432 |
+
|
| 433 |
+
|
| 434 |
+
def calc_horizontal_offset(target_img, source_img):
|
| 435 |
+
target_mask = target_img.astype(np.float32).sum(axis=-1) > 750
|
| 436 |
+
source_mask = source_img.astype(np.float32).sum(axis=-1) > 750
|
| 437 |
+
best_offset = -114514
|
| 438 |
+
for offset in range(-200, 200):
|
| 439 |
+
offset_mask = np.roll(source_mask, offset, axis=1)
|
| 440 |
+
overlap = (target_mask & offset_mask).sum()
|
| 441 |
+
if overlap > best_offset:
|
| 442 |
+
best_offset = overlap
|
| 443 |
+
best_offset_value = offset
|
| 444 |
+
return best_offset_value
|
| 445 |
+
|
| 446 |
+
|
| 447 |
+
def calc_horizontal_offset2(target_mask, source_img):
|
| 448 |
+
source_mask = source_img.astype(np.float32).sum(axis=-1) > 750
|
| 449 |
+
best_offset = -114514
|
| 450 |
+
for offset in range(-200, 200):
|
| 451 |
+
offset_mask = np.roll(source_mask, offset, axis=1)
|
| 452 |
+
overlap = (target_mask & offset_mask).sum()
|
| 453 |
+
if overlap > best_offset:
|
| 454 |
+
best_offset = overlap
|
| 455 |
+
best_offset_value = offset
|
| 456 |
+
return best_offset_value
|
| 457 |
+
|
| 458 |
+
|
| 459 |
+
@spaces.GPU
|
| 460 |
+
def get_distract_mask(generator, color_0, color_1, normal_0=None, normal_1=None, thres=0.25, ratio=0.50, outside_thres=0.10, outside_ratio=0.20):
|
| 461 |
+
distract_area = np.abs(color_0 - color_1).sum(axis=-1) > thres
|
| 462 |
+
if normal_0 is not None and normal_1 is not None:
|
| 463 |
+
distract_area |= np.abs(normal_0 - normal_1).sum(axis=-1) > thres
|
| 464 |
+
labeled_array, num_features = scipy.ndimage.label(distract_area)
|
| 465 |
+
results = []
|
| 466 |
+
|
| 467 |
+
random_sampled_points = []
|
| 468 |
+
|
| 469 |
+
for i in range(num_features + 1):
|
| 470 |
+
if np.sum(labeled_array == i) > 1000 and np.sum(labeled_array == i) < 100000:
|
| 471 |
+
results.append((i, np.sum(labeled_array == i)))
|
| 472 |
+
# random sample a point in the area
|
| 473 |
+
points = np.argwhere(labeled_array == i)
|
| 474 |
+
random_sampled_points.append(points[np.random.randint(0, points.shape[0])])
|
| 475 |
+
|
| 476 |
+
results = sorted(results, key=lambda x: x[1], reverse=True) # [1:]
|
| 477 |
+
distract_mask = np.zeros_like(distract_area)
|
| 478 |
+
distract_bbox = np.zeros_like(distract_area)
|
| 479 |
+
for i, _ in results:
|
| 480 |
+
distract_mask |= labeled_array == i
|
| 481 |
+
bbox = np.argwhere(labeled_array == i)
|
| 482 |
+
min_x, min_y = bbox.min(axis=0)
|
| 483 |
+
max_x, max_y = bbox.max(axis=0)
|
| 484 |
+
distract_bbox[min_x:max_x, min_y:max_y] = 1
|
| 485 |
+
|
| 486 |
+
points = np.array(random_sampled_points)[:, ::-1]
|
| 487 |
+
labels = np.ones(len(points), dtype=np.int32)
|
| 488 |
+
|
| 489 |
+
masks = generator.generate((color_1 * 255).astype(np.uint8))
|
| 490 |
+
|
| 491 |
+
outside_area = np.abs(color_0 - color_1).sum(axis=-1) < outside_thres
|
| 492 |
+
|
| 493 |
+
final_mask = np.zeros_like(distract_mask)
|
| 494 |
+
for iii, mask in enumerate(masks):
|
| 495 |
+
mask['segmentation'] = cv2.resize(mask['segmentation'].astype(np.float32), (1024, 1024)) > 0.5
|
| 496 |
+
intersection = np.logical_and(mask['segmentation'], distract_mask).sum()
|
| 497 |
+
total = mask['segmentation'].sum()
|
| 498 |
+
iou = intersection / total
|
| 499 |
+
outside_intersection = np.logical_and(mask['segmentation'], outside_area).sum()
|
| 500 |
+
outside_total = mask['segmentation'].sum()
|
| 501 |
+
outside_iou = outside_intersection / outside_total
|
| 502 |
+
if iou > ratio and outside_iou < outside_ratio:
|
| 503 |
+
final_mask |= mask['segmentation']
|
| 504 |
+
|
| 505 |
+
# calculate coverage
|
| 506 |
+
intersection = np.logical_and(final_mask, distract_mask).sum()
|
| 507 |
+
total = distract_mask.sum()
|
| 508 |
+
coverage = intersection / total
|
| 509 |
+
|
| 510 |
+
if coverage < 0.8:
|
| 511 |
+
# use original distract mask
|
| 512 |
+
final_mask = (distract_mask.copy() * 255).astype(np.uint8)
|
| 513 |
+
final_mask = cv2.dilate(final_mask, np.ones((3, 3), np.uint8), iterations=3)
|
| 514 |
+
labeled_array_dilate, num_features_dilate = scipy.ndimage.label(final_mask)
|
| 515 |
+
for i in range(num_features_dilate + 1):
|
| 516 |
+
if np.sum(labeled_array_dilate == i) < 200:
|
| 517 |
+
final_mask[labeled_array_dilate == i] = 255
|
| 518 |
+
|
| 519 |
+
final_mask = cv2.erode(final_mask, np.ones((3, 3), np.uint8), iterations=3)
|
| 520 |
+
final_mask = final_mask > 127
|
| 521 |
+
|
| 522 |
+
return distract_mask, distract_bbox, random_sampled_points, final_mask
|
| 523 |
+
|
| 524 |
+
|
| 525 |
+
class InferRefineAPI:
|
| 526 |
+
@spaces.GPU
|
| 527 |
+
def __init__(self, config):
|
| 528 |
+
self.sam = sam_model_registry["vit_h"](checkpoint="./ckpt/sam_vit_h_4b8939.pth").cuda()
|
| 529 |
+
self.generator = SamAutomaticMaskGenerator(
|
| 530 |
+
model=self.sam,
|
| 531 |
+
points_per_side=64,
|
| 532 |
+
pred_iou_thresh=0.80,
|
| 533 |
+
stability_score_thresh=0.92,
|
| 534 |
+
crop_n_layers=1,
|
| 535 |
+
crop_n_points_downscale_factor=2,
|
| 536 |
+
min_mask_region_area=100,
|
| 537 |
+
)
|
| 538 |
+
self.outside_ratio = 0.20
|
| 539 |
+
|
| 540 |
+
@spaces.GPU
|
| 541 |
+
def refine(self, meshes, imgs):
|
| 542 |
+
fixed_v, fixed_f, fixed_t = None, None, None
|
| 543 |
+
flow_vert, flow_vector = None, None
|
| 544 |
+
last_colors, last_normals = None, None
|
| 545 |
+
last_front_color, last_front_normal = None, None
|
| 546 |
+
distract_mask = None
|
| 547 |
+
|
| 548 |
+
mv, proj = make_star_cameras_orthographic(8, 1, r=1.2)
|
| 549 |
+
mv = mv[[4, 3, 2, 0, 6, 5]]
|
| 550 |
+
renderer = NormalsRenderer(mv,proj,(1024,1024))
|
| 551 |
+
|
| 552 |
+
results = []
|
| 553 |
+
|
| 554 |
+
for name_idx, level in zip([2, 0, 1], [2, 1, 0]):
|
| 555 |
+
mesh = trimesh.load(meshes[name_idx])
|
| 556 |
+
new_mesh = mesh.split(only_watertight=False)
|
| 557 |
+
new_mesh = [ j for j in new_mesh if len(j.vertices) >= 300 ]
|
| 558 |
+
mesh = trimesh.Scene(new_mesh).dump(concatenate=True)
|
| 559 |
+
mesh_v, mesh_f = mesh.vertices, mesh.faces
|
| 560 |
+
|
| 561 |
+
if last_colors is None:
|
| 562 |
+
images = renderer.render(
|
| 563 |
+
torch.tensor(mesh_v, device='cuda').float(),
|
| 564 |
+
torch.ones_like(torch.from_numpy(mesh_v), device='cuda').float(),
|
| 565 |
+
torch.tensor(mesh_f, device='cuda'),
|
| 566 |
+
)
|
| 567 |
+
mask = (images[..., 3] < 0.9).cpu().numpy()
|
| 568 |
+
|
| 569 |
+
colors, normals = [], []
|
| 570 |
+
for i in range(6):
|
| 571 |
+
color = np.array(imgs[level]['images'][i])
|
| 572 |
+
normal = np.array(imgs[level]['normals'][i])
|
| 573 |
+
|
| 574 |
+
if last_colors is not None:
|
| 575 |
+
offset = calc_horizontal_offset(np.array(last_colors[i]), color)
|
| 576 |
+
# print('offset', i, offset)
|
| 577 |
+
else:
|
| 578 |
+
offset = calc_horizontal_offset2(mask[i], color)
|
| 579 |
+
# print('init offset', i, offset)
|
| 580 |
+
|
| 581 |
+
if offset != 0:
|
| 582 |
+
color = np.roll(color, offset, axis=1)
|
| 583 |
+
normal = np.roll(normal, offset, axis=1)
|
| 584 |
+
|
| 585 |
+
color = Image.fromarray(color)
|
| 586 |
+
normal = Image.fromarray(normal)
|
| 587 |
+
colors.append(color)
|
| 588 |
+
normals.append(normal)
|
| 589 |
+
|
| 590 |
+
if last_front_color is not None and level == 0:
|
| 591 |
+
original_mask, distract_bbox, _, distract_mask = get_distract_mask(self.generator, last_front_color, np.array(colors[0]).astype(np.float32) / 255.0, outside_ratio=self.outside_ratio)
|
| 592 |
+
else:
|
| 593 |
+
distract_mask = None
|
| 594 |
+
distract_bbox = None
|
| 595 |
+
|
| 596 |
+
last_front_color = np.array(colors[0]).astype(np.float32) / 255.0
|
| 597 |
+
last_front_normal = np.array(normals[0]).astype(np.float32) / 255.0
|
| 598 |
+
|
| 599 |
+
if last_colors is None:
|
| 600 |
+
from copy import deepcopy
|
| 601 |
+
last_colors, last_normals = deepcopy(colors), deepcopy(normals)
|
| 602 |
+
|
| 603 |
+
# my mesh flow weight by nearest vertexs
|
| 604 |
+
if fixed_v is not None and fixed_f is not None and level == 1:
|
| 605 |
+
t = trimesh.Trimesh(vertices=mesh_v, faces=mesh_f)
|
| 606 |
+
|
| 607 |
+
fixed_v_cpu = fixed_v.cpu().numpy()
|
| 608 |
+
kdtree_anchor = KDTree(fixed_v_cpu)
|
| 609 |
+
kdtree_mesh_v = KDTree(mesh_v)
|
| 610 |
+
_, idx_anchor = kdtree_anchor.query(mesh_v, k=1)
|
| 611 |
+
_, idx_mesh_v = kdtree_mesh_v.query(mesh_v, k=25)
|
| 612 |
+
idx_anchor = idx_anchor.squeeze()
|
| 613 |
+
neighbors = torch.tensor(mesh_v).cuda()[idx_mesh_v] # V, 25, 3
|
| 614 |
+
# calculate the distances neighbors [V, 25, 3]; mesh_v [V, 3] -> [V, 25]
|
| 615 |
+
neighbor_dists = torch.norm(neighbors - torch.tensor(mesh_v).cuda()[:, None], dim=-1)
|
| 616 |
+
neighbor_dists[neighbor_dists > 0.06] = 114514.
|
| 617 |
+
neighbor_weights = torch.exp(-neighbor_dists * 1.)
|
| 618 |
+
neighbor_weights = neighbor_weights / neighbor_weights.sum(dim=1, keepdim=True)
|
| 619 |
+
anchors = fixed_v[idx_anchor] # V, 3
|
| 620 |
+
anchor_normals = calc_vertex_normals(fixed_v, fixed_f)[idx_anchor] # V, 3
|
| 621 |
+
dis_anchor = torch.clamp(((anchors - torch.tensor(mesh_v).cuda()) * anchor_normals).sum(-1), min=0) + 0.01
|
| 622 |
+
vec_anchor = dis_anchor[:, None] * anchor_normals # V, 3
|
| 623 |
+
vec_anchor = vec_anchor[idx_mesh_v] # V, 25, 3
|
| 624 |
+
weighted_vec_anchor = (vec_anchor * neighbor_weights[:, :, None]).sum(1) # V, 3
|
| 625 |
+
mesh_v += weighted_vec_anchor.cpu().numpy()
|
| 626 |
+
|
| 627 |
+
t = trimesh.Trimesh(vertices=mesh_v, faces=mesh_f)
|
| 628 |
+
|
| 629 |
+
mesh_v = torch.tensor(mesh_v, device='cuda', dtype=torch.float32)
|
| 630 |
+
mesh_f = torch.tensor(mesh_f, device='cuda')
|
| 631 |
+
|
| 632 |
+
new_mesh, simp_v, simp_f = geo_refine(mesh_v, mesh_f, colors, normals, fixed_v=fixed_v, fixed_f=fixed_f, distract_mask=distract_mask, distract_bbox=distract_bbox)
|
| 633 |
+
|
| 634 |
+
# my mesh flow weight by nearest vertexs
|
| 635 |
+
try:
|
| 636 |
+
if fixed_v is not None and fixed_f is not None and level != 0:
|
| 637 |
+
new_mesh_v = new_mesh.verts_packed().cpu().numpy()
|
| 638 |
+
|
| 639 |
+
fixed_v_cpu = fixed_v.cpu().numpy()
|
| 640 |
+
kdtree_anchor = KDTree(fixed_v_cpu)
|
| 641 |
+
kdtree_mesh_v = KDTree(new_mesh_v)
|
| 642 |
+
_, idx_anchor = kdtree_anchor.query(new_mesh_v, k=1)
|
| 643 |
+
_, idx_mesh_v = kdtree_mesh_v.query(new_mesh_v, k=25)
|
| 644 |
+
idx_anchor = idx_anchor.squeeze()
|
| 645 |
+
neighbors = torch.tensor(new_mesh_v).cuda()[idx_mesh_v] # V, 25, 3
|
| 646 |
+
# calculate the distances neighbors [V, 25, 3]; new_mesh_v [V, 3] -> [V, 25]
|
| 647 |
+
neighbor_dists = torch.norm(neighbors - torch.tensor(new_mesh_v).cuda()[:, None], dim=-1)
|
| 648 |
+
neighbor_dists[neighbor_dists > 0.06] = 114514.
|
| 649 |
+
neighbor_weights = torch.exp(-neighbor_dists * 1.)
|
| 650 |
+
neighbor_weights = neighbor_weights / neighbor_weights.sum(dim=1, keepdim=True)
|
| 651 |
+
anchors = fixed_v[idx_anchor] # V, 3
|
| 652 |
+
anchor_normals = calc_vertex_normals(fixed_v, fixed_f)[idx_anchor] # V, 3
|
| 653 |
+
dis_anchor = torch.clamp(((anchors - torch.tensor(new_mesh_v).cuda()) * anchor_normals).sum(-1), min=0) + 0.01
|
| 654 |
+
vec_anchor = dis_anchor[:, None] * anchor_normals # V, 3
|
| 655 |
+
vec_anchor = vec_anchor[idx_mesh_v] # V, 25, 3
|
| 656 |
+
weighted_vec_anchor = (vec_anchor * neighbor_weights[:, :, None]).sum(1) # V, 3
|
| 657 |
+
new_mesh_v += weighted_vec_anchor.cpu().numpy()
|
| 658 |
+
|
| 659 |
+
# replace new_mesh verts with new_mesh_v
|
| 660 |
+
new_mesh = Meshes(verts=[torch.tensor(new_mesh_v, device='cuda')], faces=new_mesh.faces_list(), textures=new_mesh.textures)
|
| 661 |
+
|
| 662 |
+
except Exception as e:
|
| 663 |
+
pass
|
| 664 |
+
|
| 665 |
+
notsimp_v, notsimp_f, notsimp_t = new_mesh.verts_packed(), new_mesh.faces_packed(), new_mesh.textures.verts_features_packed()
|
| 666 |
+
|
| 667 |
+
if fixed_v is None:
|
| 668 |
+
fixed_v, fixed_f = simp_v, simp_f
|
| 669 |
+
complete_v, complete_f, complete_t = notsimp_v, notsimp_f, notsimp_t
|
| 670 |
+
else:
|
| 671 |
+
fixed_f = torch.cat([fixed_f, simp_f + fixed_v.shape[0]], dim=0)
|
| 672 |
+
fixed_v = torch.cat([fixed_v, simp_v], dim=0)
|
| 673 |
+
|
| 674 |
+
complete_f = torch.cat([complete_f, notsimp_f + complete_v.shape[0]], dim=0)
|
| 675 |
+
complete_v = torch.cat([complete_v, notsimp_v], dim=0)
|
| 676 |
+
complete_t = torch.cat([complete_t, notsimp_t], dim=0)
|
| 677 |
+
|
| 678 |
+
if level == 2:
|
| 679 |
+
new_mesh = Meshes(verts=[new_mesh.verts_packed()], faces=[new_mesh.faces_packed()], textures=pytorch3d.renderer.mesh.textures.TexturesVertex(verts_features=[torch.ones_like(new_mesh.textures.verts_features_packed(), device=new_mesh.verts_packed().device)*0.5]))
|
| 680 |
+
|
| 681 |
+
save_py3dmesh_with_trimesh_fast(new_mesh, meshes[name_idx].replace('.obj', '_refined.obj'), apply_sRGB_to_LinearRGB=False)
|
| 682 |
+
results.append(meshes[name_idx].replace('.obj', '_refined.obj'))
|
| 683 |
+
|
| 684 |
+
# save whole mesh
|
| 685 |
+
save_py3dmesh_with_trimesh_fast(Meshes(verts=[complete_v], faces=[complete_f], textures=pytorch3d.renderer.mesh.textures.TexturesVertex(verts_features=[complete_t])), meshes[name_idx].replace('.obj', '_refined_whole.obj'), apply_sRGB_to_LinearRGB=False)
|
| 686 |
+
results.append(meshes[name_idx].replace('.obj', '_refined_whole.obj'))
|
| 687 |
+
|
| 688 |
+
return results
|
| 689 |
+
|
| 690 |
+
|
| 691 |
+
class InferSlrmAPI:
|
| 692 |
+
@spaces.GPU
|
| 693 |
+
def __init__(self, config):
|
| 694 |
+
self.config_path = config['config_path']
|
| 695 |
+
self.config = OmegaConf.load(self.config_path)
|
| 696 |
+
self.config_name = os.path.basename(self.config_path).replace('.yaml', '')
|
| 697 |
+
self.model_config = self.config.model_config
|
| 698 |
+
self.infer_config = self.config.infer_config
|
| 699 |
+
self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
| 700 |
+
self.model = instantiate_from_config(self.model_config)
|
| 701 |
+
state_dict = torch.load(self.infer_config.model_path, map_location='cpu')
|
| 702 |
+
self.model.load_state_dict(state_dict, strict=False)
|
| 703 |
+
self.model = self.model.to(self.device)
|
| 704 |
+
self.model.init_flexicubes_geometry(self.device, fovy=30.0, is_ortho=self.model.is_ortho)
|
| 705 |
+
self.model = self.model.eval()
|
| 706 |
+
|
| 707 |
+
@spaces.GPU
|
| 708 |
+
def gen(self, imgs):
|
| 709 |
+
imgs = [ cv2.imread(img[0])[:, :, ::-1] for img in imgs ]
|
| 710 |
+
imgs = np.stack(imgs, axis=0).astype(np.float32) / 255.0
|
| 711 |
+
imgs = torch.from_numpy(np.array(imgs)).permute(0, 3, 1, 2).contiguous().float() # (6, 3, 1024, 1024)
|
| 712 |
+
mesh_glb_fpaths = self.make3d(imgs)
|
| 713 |
+
return mesh_glb_fpaths[1:4] + mesh_glb_fpaths[0:1]
|
| 714 |
+
|
| 715 |
+
@spaces.GPU
|
| 716 |
+
def make3d(self, images):
|
| 717 |
+
input_cameras = torch.tensor(np.load('slrm/cameras.npy')).to(device)
|
| 718 |
+
|
| 719 |
+
images = images.unsqueeze(0).to(device)
|
| 720 |
+
images = v2.functional.resize(images, (320, 320), interpolation=3, antialias=True).clamp(0, 1)
|
| 721 |
+
|
| 722 |
+
mesh_fpath = tempfile.NamedTemporaryFile(suffix=f".obj", delete=False).name
|
| 723 |
+
print(mesh_fpath)
|
| 724 |
+
mesh_basename = os.path.basename(mesh_fpath).split('.')[0]
|
| 725 |
+
mesh_dirname = os.path.dirname(mesh_fpath)
|
| 726 |
+
|
| 727 |
+
with torch.no_grad():
|
| 728 |
+
# get triplane
|
| 729 |
+
planes = self.model.forward_planes(images, input_cameras.float())
|
| 730 |
+
|
| 731 |
+
# get mesh
|
| 732 |
+
mesh_glb_fpaths = []
|
| 733 |
+
for j in range(4):
|
| 734 |
+
mesh_glb_fpath = self.make_mesh(mesh_fpath.replace(mesh_fpath[-4:], f'_{j}{mesh_fpath[-4:]}'), planes, level=[0, 3, 4, 2][j])
|
| 735 |
+
mesh_glb_fpaths.append(mesh_glb_fpath)
|
| 736 |
+
|
| 737 |
+
return mesh_glb_fpaths
|
| 738 |
+
|
| 739 |
+
@spaces.GPU
|
| 740 |
+
def make_mesh(self, mesh_fpath, planes, level=None):
|
| 741 |
+
mesh_basename = os.path.basename(mesh_fpath).split('.')[0]
|
| 742 |
+
mesh_dirname = os.path.dirname(mesh_fpath)
|
| 743 |
+
mesh_glb_fpath = os.path.join(mesh_dirname, f"{mesh_basename}.glb")
|
| 744 |
+
|
| 745 |
+
with torch.no_grad():
|
| 746 |
+
# get mesh
|
| 747 |
+
mesh_out = self.model.extract_mesh(
|
| 748 |
+
planes,
|
| 749 |
+
use_texture_map=False,
|
| 750 |
+
levels=torch.tensor([level]).to(device),
|
| 751 |
+
**self.infer_config,
|
| 752 |
+
)
|
| 753 |
+
|
| 754 |
+
vertices, faces, vertex_colors = mesh_out
|
| 755 |
+
vertices = vertices[:, [1, 2, 0]]
|
| 756 |
+
|
| 757 |
+
if level == 2:
|
| 758 |
+
# fill all vertex_colors with 127
|
| 759 |
+
vertex_colors = np.ones_like(vertex_colors) * 127
|
| 760 |
+
|
| 761 |
+
save_obj(vertices, faces, vertex_colors, mesh_fpath)
|
| 762 |
+
|
| 763 |
+
return mesh_fpath
|
| 764 |
+
|
| 765 |
+
class InferMultiviewAPI:
|
| 766 |
+
def __init__(self, config):
|
| 767 |
+
parser = argparse.ArgumentParser()
|
| 768 |
+
parser.add_argument("--seed", type=int, default=42)
|
| 769 |
+
parser.add_argument("--num_views", type=int, default=6)
|
| 770 |
+
parser.add_argument("--num_levels", type=int, default=3)
|
| 771 |
+
parser.add_argument("--pretrained_path", type=str, default='./ckpt/StdGEN-multiview-1024')
|
| 772 |
+
parser.add_argument("--height", type=int, default=1024)
|
| 773 |
+
parser.add_argument("--width", type=int, default=576)
|
| 774 |
+
self.cfg = parser.parse_args()
|
| 775 |
+
self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
| 776 |
+
self.pipeline = load_multiview_pipeline(self.cfg)
|
| 777 |
+
self.results = {}
|
| 778 |
+
if torch.cuda.is_available():
|
| 779 |
+
self.pipeline.to(device)
|
| 780 |
+
|
| 781 |
+
self.image_transforms = [transforms.Resize(int(max(self.cfg.height, self.cfg.width))),
|
| 782 |
+
transforms.CenterCrop((self.cfg.height, self.cfg.width)),
|
| 783 |
+
transforms.ToTensor(),
|
| 784 |
+
transforms.Lambda(lambda x: x * 2. - 1),
|
| 785 |
+
]
|
| 786 |
+
self.image_transforms = transforms.Compose(self.image_transforms)
|
| 787 |
+
|
| 788 |
+
prompt_embeds_path = './multiview/fixed_prompt_embeds_6view'
|
| 789 |
+
self.normal_text_embeds = torch.load(f'{prompt_embeds_path}/normal_embeds.pt')
|
| 790 |
+
self.color_text_embeds = torch.load(f'{prompt_embeds_path}/clr_embeds.pt')
|
| 791 |
+
self.total_views = self.cfg.num_views
|
| 792 |
+
|
| 793 |
+
|
| 794 |
+
def process_im(self, im):
|
| 795 |
+
im = self.image_transforms(im)
|
| 796 |
+
return im
|
| 797 |
+
|
| 798 |
+
def gen(self, img, seed, num_levels):
|
| 799 |
+
set_seed(seed)
|
| 800 |
+
data = {}
|
| 801 |
+
|
| 802 |
+
cond_im_rgb = self.process_im(img)
|
| 803 |
+
cond_im_rgb = torch.stack([cond_im_rgb] * self.total_views, dim=0)
|
| 804 |
+
data["image_cond_rgb"] = cond_im_rgb[None, ...]
|
| 805 |
+
data["normal_prompt_embeddings"] = self.normal_text_embeds[None, ...]
|
| 806 |
+
data["color_prompt_embeddings"] = self.color_text_embeds[None, ...]
|
| 807 |
+
|
| 808 |
+
results = run_multiview_infer(data, self.pipeline, self.cfg, num_levels=num_levels)
|
| 809 |
+
for k in results:
|
| 810 |
+
self.results[k] = results[k]
|
| 811 |
+
return results
|
| 812 |
+
|
| 813 |
+
|
| 814 |
+
class InferCanonicalAPI:
|
| 815 |
+
def __init__(self, config):
|
| 816 |
+
self.config = config
|
| 817 |
+
self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
| 818 |
+
|
| 819 |
+
self.config_path = config['config_path']
|
| 820 |
+
self.loaded_config = OmegaConf.load(self.config_path)
|
| 821 |
+
|
| 822 |
+
self.setup(**self.loaded_config)
|
| 823 |
+
|
| 824 |
+
def setup(self,
|
| 825 |
+
validation: Dict,
|
| 826 |
+
pretrained_model_path: str,
|
| 827 |
+
local_crossattn: bool = True,
|
| 828 |
+
unet_from_pretrained_kwargs=None,
|
| 829 |
+
unet_condition_type=None,
|
| 830 |
+
use_noise=True,
|
| 831 |
+
noise_d=256,
|
| 832 |
+
timestep: int = 40,
|
| 833 |
+
width_input: int = 640,
|
| 834 |
+
height_input: int = 1024,
|
| 835 |
+
):
|
| 836 |
+
self.width_input = width_input
|
| 837 |
+
self.height_input = height_input
|
| 838 |
+
self.timestep = timestep
|
| 839 |
+
self.use_noise = use_noise
|
| 840 |
+
self.noise_d = noise_d
|
| 841 |
+
self.validation = validation
|
| 842 |
+
self.unet_condition_type = unet_condition_type
|
| 843 |
+
self.pretrained_model_path = pretrained_model_path
|
| 844 |
+
|
| 845 |
+
self.tokenizer = CLIPTokenizer.from_pretrained(pretrained_model_path, subfolder="tokenizer")
|
| 846 |
+
self.text_encoder = CLIPTextModel.from_pretrained(pretrained_model_path, subfolder="text_encoder")
|
| 847 |
+
self.image_encoder = CLIPVisionModelWithProjection.from_pretrained(pretrained_model_path, subfolder="image_encoder")
|
| 848 |
+
self.feature_extractor = CLIPImageProcessor()
|
| 849 |
+
self.vae = AutoencoderKL.from_pretrained(pretrained_model_path, subfolder="vae")
|
| 850 |
+
self.unet = UNetMV2DConditionModel.from_pretrained_2d(pretrained_model_path, subfolder="unet", local_crossattn=local_crossattn, **unet_from_pretrained_kwargs)
|
| 851 |
+
self.ref_unet = UNetMV2DRefModel.from_pretrained_2d(pretrained_model_path, subfolder="ref_unet", local_crossattn=local_crossattn, **unet_from_pretrained_kwargs)
|
| 852 |
+
|
| 853 |
+
self.text_encoder.to(device, dtype=weight_dtype)
|
| 854 |
+
self.image_encoder.to(device, dtype=weight_dtype)
|
| 855 |
+
self.vae.to(device, dtype=weight_dtype)
|
| 856 |
+
self.ref_unet.to(device, dtype=weight_dtype)
|
| 857 |
+
self.unet.to(device, dtype=weight_dtype)
|
| 858 |
+
|
| 859 |
+
self.vae.requires_grad_(False)
|
| 860 |
+
self.ref_unet.requires_grad_(False)
|
| 861 |
+
self.unet.requires_grad_(False)
|
| 862 |
+
|
| 863 |
+
self.noise_scheduler = DDIMScheduler.from_pretrained(pretrained_model_path, subfolder="scheduler-zerosnr")
|
| 864 |
+
self.validation_pipeline = CanonicalizationPipeline(
|
| 865 |
+
vae=self.vae, text_encoder=self.text_encoder, tokenizer=self.tokenizer, unet=self.unet, ref_unet=self.ref_unet,feature_extractor=self.feature_extractor,image_encoder=self.image_encoder,
|
| 866 |
+
scheduler=self.noise_scheduler
|
| 867 |
+
)
|
| 868 |
+
self.validation_pipeline.set_progress_bar_config(disable=True)
|
| 869 |
+
|
| 870 |
+
def canonicalize(self, image, seed):
|
| 871 |
+
return inference(
|
| 872 |
+
self.validation_pipeline, image, self.vae, self.feature_extractor, self.image_encoder, self.unet, self.ref_unet, self.tokenizer, self.text_encoder,
|
| 873 |
+
self.pretrained_model_path, self.validation, self.width_input, self.height_input, self.unet_condition_type,
|
| 874 |
+
use_noise=self.use_noise, noise_d=self.noise_d, crop=True, seed=seed, timestep=self.timestep
|
| 875 |
+
)
|
| 876 |
+
|
| 877 |
+
def gen(self, img_input, seed=0):
|
| 878 |
+
if np.array(img_input).shape[-1] == 4 and np.array(img_input)[..., 3].min() == 255:
|
| 879 |
+
# convert to RGB
|
| 880 |
+
img_input = img_input.convert("RGB")
|
| 881 |
+
img_output = self.canonicalize(img_input, seed)
|
| 882 |
+
|
| 883 |
+
max_dim = max(img_output.width, img_output.height)
|
| 884 |
+
new_image = Image.new("RGBA", (max_dim, max_dim))
|
| 885 |
+
left = (max_dim - img_output.width) // 2
|
| 886 |
+
top = (max_dim - img_output.height) // 2
|
| 887 |
+
new_image.paste(img_output, (left, top))
|
| 888 |
+
|
| 889 |
+
return new_image
|