Spaces:
Paused
Paused
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,341 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import imageio
|
3 |
+
import numpy as np
|
4 |
+
import torch
|
5 |
+
import rembg
|
6 |
+
from PIL import Image
|
7 |
+
from torchvision.transforms import v2
|
8 |
+
from pytorch_lightning import seed_everything
|
9 |
+
from omegaconf import OmegaConf
|
10 |
+
from einops import rearrange, repeat
|
11 |
+
from tqdm import tqdm
|
12 |
+
from diffusers import DiffusionPipeline, EulerAncestralDiscreteScheduler
|
13 |
+
|
14 |
+
from src.utils.train_util import instantiate_from_config
|
15 |
+
from src.utils.camera_util import (
|
16 |
+
FOV_to_intrinsics,
|
17 |
+
get_zero123plus_input_cameras,
|
18 |
+
get_circular_camera_poses,
|
19 |
+
)
|
20 |
+
from src.utils.mesh_util import save_obj
|
21 |
+
from src.utils.infer_util import remove_background, resize_foreground, images_to_video
|
22 |
+
|
23 |
+
import tempfile
|
24 |
+
from functools import partial
|
25 |
+
|
26 |
+
from huggingface_hub import hf_hub_download
|
27 |
+
|
28 |
+
import gradio as gr
|
29 |
+
import spaces
|
30 |
+
|
31 |
+
|
32 |
+
def get_render_cameras(batch_size=1, M=120, radius=2.5, elevation=10.0, is_flexicubes=False):
|
33 |
+
"""
|
34 |
+
Get the rendering camera parameters.
|
35 |
+
"""
|
36 |
+
c2ws = get_circular_camera_poses(M=M, radius=radius, elevation=elevation)
|
37 |
+
if is_flexicubes:
|
38 |
+
cameras = torch.linalg.inv(c2ws)
|
39 |
+
cameras = cameras.unsqueeze(0).repeat(batch_size, 1, 1, 1)
|
40 |
+
else:
|
41 |
+
extrinsics = c2ws.flatten(-2)
|
42 |
+
intrinsics = FOV_to_intrinsics(50.0).unsqueeze(0).repeat(M, 1, 1).float().flatten(-2)
|
43 |
+
cameras = torch.cat([extrinsics, intrinsics], dim=-1)
|
44 |
+
cameras = cameras.unsqueeze(0).repeat(batch_size, 1, 1)
|
45 |
+
return cameras
|
46 |
+
|
47 |
+
|
48 |
+
def images_to_video(images, output_path, fps=30):
|
49 |
+
# images: (N, C, H, W)
|
50 |
+
os.makedirs(os.path.dirname(output_path), exist_ok=True)
|
51 |
+
frames = []
|
52 |
+
for i in range(images.shape[0]):
|
53 |
+
frame = (images[i].permute(1, 2, 0).cpu().numpy() * 255).astype(np.uint8).clip(0, 255)
|
54 |
+
assert frame.shape[0] == images.shape[2] and frame.shape[1] == images.shape[3], \
|
55 |
+
f"Frame shape mismatch: {frame.shape} vs {images.shape}"
|
56 |
+
assert frame.min() >= 0 and frame.max() <= 255, \
|
57 |
+
f"Frame value out of range: {frame.min()} ~ {frame.max()}"
|
58 |
+
frames.append(frame)
|
59 |
+
imageio.mimwrite(output_path, np.stack(frames), fps=fps, codec='h264')
|
60 |
+
|
61 |
+
|
62 |
+
###############################################################################
|
63 |
+
# Configuration.
|
64 |
+
###############################################################################
|
65 |
+
|
66 |
+
config_path = 'configs/instant-mesh-large.yaml'
|
67 |
+
config = OmegaConf.load(config_path)
|
68 |
+
config_name = os.path.basename(config_path).replace('.yaml', '')
|
69 |
+
model_config = config.model_config
|
70 |
+
infer_config = config.infer_config
|
71 |
+
|
72 |
+
IS_FLEXICUBES = True if config_name.startswith('instant-mesh') else False
|
73 |
+
|
74 |
+
device = torch.device('cuda')
|
75 |
+
|
76 |
+
# load diffusion model
|
77 |
+
print('Loading diffusion model ...')
|
78 |
+
pipeline = DiffusionPipeline.from_pretrained(
|
79 |
+
"sudo-ai/zero123plus-v1.2",
|
80 |
+
custom_pipeline="zero123plus",
|
81 |
+
torch_dtype=torch.float16,
|
82 |
+
)
|
83 |
+
pipeline.scheduler = EulerAncestralDiscreteScheduler.from_config(
|
84 |
+
pipeline.scheduler.config, timestep_spacing='trailing'
|
85 |
+
)
|
86 |
+
|
87 |
+
# load custom white-background UNet
|
88 |
+
unet_ckpt_path = hf_hub_download(repo_id="TencentARC/InstantMesh", filename="diffusion_pytorch_model.bin", repo_type="model")
|
89 |
+
state_dict = torch.load(unet_ckpt_path, map_location='cpu')
|
90 |
+
pipeline.unet.load_state_dict(state_dict, strict=True)
|
91 |
+
|
92 |
+
pipeline = pipeline.to(device)
|
93 |
+
|
94 |
+
# load reconstruction model
|
95 |
+
print('Loading reconstruction model ...')
|
96 |
+
model_ckpt_path = hf_hub_download(repo_id="TencentARC/InstantMesh", filename="instant_mesh_large.ckpt", repo_type="model")
|
97 |
+
model = instantiate_from_config(model_config)
|
98 |
+
state_dict = torch.load(model_ckpt_path, map_location='cpu')['state_dict']
|
99 |
+
state_dict = {k[14:]: v for k, v in state_dict.items() if k.startswith('lrm_generator.') and 'source_camera' not in k}
|
100 |
+
model.load_state_dict(state_dict, strict=True)
|
101 |
+
|
102 |
+
model = model.to(device)
|
103 |
+
if IS_FLEXICUBES:
|
104 |
+
model.init_flexicubes_geometry(device)
|
105 |
+
model = model.eval()
|
106 |
+
|
107 |
+
print('Loading Finished!')
|
108 |
+
|
109 |
+
|
110 |
+
def check_input_image(input_image):
|
111 |
+
if input_image is None:
|
112 |
+
raise gr.Error("No image uploaded!")
|
113 |
+
|
114 |
+
|
115 |
+
def preprocess(input_image, do_remove_background):
|
116 |
+
|
117 |
+
rembg_session = rembg.new_session() if do_remove_background else None
|
118 |
+
|
119 |
+
if do_remove_background:
|
120 |
+
input_image = remove_background(input_image, rembg_session)
|
121 |
+
input_image = resize_foreground(input_image, 0.85)
|
122 |
+
|
123 |
+
return input_image
|
124 |
+
|
125 |
+
|
126 |
+
def generate_mvs(input_image, sample_steps, sample_seed):
|
127 |
+
|
128 |
+
seed_everything(sample_seed)
|
129 |
+
|
130 |
+
# sampling
|
131 |
+
z123_image = pipeline(
|
132 |
+
input_image,
|
133 |
+
num_inference_steps=sample_steps
|
134 |
+
).images[0]
|
135 |
+
|
136 |
+
show_image = np.asarray(z123_image, dtype=np.uint8)
|
137 |
+
show_image = torch.from_numpy(show_image) # (960, 640, 3)
|
138 |
+
show_image = rearrange(show_image, '(n h) (m w) c -> (m h) (n w) c', n=3, m=2)
|
139 |
+
show_image = Image.fromarray(show_image.numpy())
|
140 |
+
|
141 |
+
return z123_image, show_image
|
142 |
+
|
143 |
+
def make_mesh(mesh_fpath, planes):
|
144 |
+
|
145 |
+
mesh_basename = os.path.basename(mesh_fpath).split('.')[0]
|
146 |
+
mesh_dirname = os.path.dirname(mesh_fpath)
|
147 |
+
mesh_vis_fpath = os.path.join(mesh_dirname, f"{mesh_basename}.glb")
|
148 |
+
|
149 |
+
with torch.no_grad():
|
150 |
+
|
151 |
+
# get mesh
|
152 |
+
mesh_out = model.extract_mesh(
|
153 |
+
planes,
|
154 |
+
use_texture_map=False,
|
155 |
+
**infer_config,
|
156 |
+
)
|
157 |
+
|
158 |
+
vertices, faces, vertex_colors = mesh_out
|
159 |
+
vertices = vertices[:, [0, 2, 1]]
|
160 |
+
vertices[:, -1] *= -1
|
161 |
+
|
162 |
+
save_obj(vertices, faces, vertex_colors, mesh_fpath)
|
163 |
+
|
164 |
+
print(f"Mesh saved to {mesh_fpath}")
|
165 |
+
|
166 |
+
return mesh_fpath
|
167 |
+
|
168 |
+
@spaces.GPU
|
169 |
+
def make3d(input_image, sample_steps, sample_seed):
|
170 |
+
|
171 |
+
images, show_images = generate_mvs(input_image, sample_steps, sample_seed)
|
172 |
+
|
173 |
+
images = np.asarray(images, dtype=np.float32) / 255.0
|
174 |
+
images = torch.from_numpy(images).permute(2, 0, 1).contiguous().float() # (3, 960, 640)
|
175 |
+
images = rearrange(images, 'c (n h) (m w) -> (n m) c h w', n=3, m=2) # (6, 3, 320, 320)
|
176 |
+
|
177 |
+
input_cameras = get_zero123plus_input_cameras(batch_size=1, radius=2.5).to(device)
|
178 |
+
render_cameras = get_render_cameras(batch_size=1, radius=2.5, is_flexicubes=IS_FLEXICUBES).to(device)
|
179 |
+
|
180 |
+
images = images.unsqueeze(0).to(device)
|
181 |
+
images = v2.functional.resize(images, (320, 320), interpolation=3, antialias=True).clamp(0, 1)
|
182 |
+
|
183 |
+
mesh_fpath = tempfile.NamedTemporaryFile(suffix=f".obj", delete=False).name
|
184 |
+
print(mesh_fpath)
|
185 |
+
mesh_basename = os.path.basename(mesh_fpath).split('.')[0]
|
186 |
+
mesh_dirname = os.path.dirname(mesh_fpath)
|
187 |
+
video_fpath = os.path.join(mesh_dirname, f"{mesh_basename}.mp4")
|
188 |
+
|
189 |
+
with torch.no_grad():
|
190 |
+
# get triplane
|
191 |
+
planes = model.forward_planes(images, input_cameras)
|
192 |
+
|
193 |
+
# get video
|
194 |
+
chunk_size = 20 if IS_FLEXICUBES else 1
|
195 |
+
render_size = 384
|
196 |
+
|
197 |
+
frames = []
|
198 |
+
for i in tqdm(range(0, render_cameras.shape[1], chunk_size)):
|
199 |
+
if IS_FLEXICUBES:
|
200 |
+
frame = model.forward_geometry(
|
201 |
+
planes,
|
202 |
+
render_cameras[:, i:i+chunk_size],
|
203 |
+
render_size=render_size,
|
204 |
+
)['img']
|
205 |
+
else:
|
206 |
+
frame = model.synthesizer(
|
207 |
+
planes,
|
208 |
+
cameras=render_cameras[:, i:i+chunk_size],
|
209 |
+
render_size=render_size,
|
210 |
+
)['images_rgb']
|
211 |
+
frames.append(frame)
|
212 |
+
frames = torch.cat(frames, dim=1)
|
213 |
+
|
214 |
+
images_to_video(
|
215 |
+
frames[0],
|
216 |
+
video_fpath,
|
217 |
+
fps=30,
|
218 |
+
)
|
219 |
+
|
220 |
+
print(f"Video saved to {video_fpath}")
|
221 |
+
|
222 |
+
mesh_fpath = make_mesh(mesh_fpath, planes)
|
223 |
+
|
224 |
+
return video_fpath, mesh_fpath, show_images
|
225 |
+
|
226 |
+
|
227 |
+
_HEADER_ = '''
|
228 |
+
<h2><b>Official 🤗 Gradio demo for</b>
|
229 |
+
<a href='https://github.com/TencentARC/InstantMesh' target='_blank'>
|
230 |
+
<b>InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models</b>
|
231 |
+
</a>.
|
232 |
+
</h2>
|
233 |
+
'''
|
234 |
+
|
235 |
+
_LINKS_ = '''
|
236 |
+
<h3>Code is available at <a href='https://github.com/TencentARC/InstantMesh' target='_blank'>GitHub</a></h3>
|
237 |
+
<h3>Report is available at <a href='https://arxiv.org/abs/2404.07191' target='_blank'>ArXiv</a></h3>
|
238 |
+
'''
|
239 |
+
|
240 |
+
_CITE_ = r"""
|
241 |
+
```bibtex
|
242 |
+
@article{xu2024instantmesh,
|
243 |
+
title={InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models},
|
244 |
+
author={Xu, Jiale and Cheng, Weihao and Gao, Yiming and Wang, Xintao and Gao, Shenghua and Shan, Ying},
|
245 |
+
journal={arXiv preprint arXiv:2404.07191},
|
246 |
+
year={2024}
|
247 |
+
}
|
248 |
+
```
|
249 |
+
"""
|
250 |
+
|
251 |
+
|
252 |
+
with gr.Blocks() as demo:
|
253 |
+
gr.Markdown(_HEADER_)
|
254 |
+
with gr.Row(variant="panel"):
|
255 |
+
with gr.Column():
|
256 |
+
with gr.Row():
|
257 |
+
input_image = gr.Image(
|
258 |
+
label="Input Image",
|
259 |
+
image_mode="RGBA",
|
260 |
+
sources="upload",
|
261 |
+
width=256,
|
262 |
+
height=256,
|
263 |
+
type="pil",
|
264 |
+
elem_id="content_image",
|
265 |
+
)
|
266 |
+
processed_image = gr.Image(
|
267 |
+
label="Processed Image",
|
268 |
+
image_mode="RGBA",
|
269 |
+
width=256,
|
270 |
+
height=256,
|
271 |
+
type="pil",
|
272 |
+
interactive=False
|
273 |
+
)
|
274 |
+
with gr.Row():
|
275 |
+
with gr.Group():
|
276 |
+
do_remove_background = gr.Checkbox(
|
277 |
+
label="Remove Background", value=True
|
278 |
+
)
|
279 |
+
sample_seed = gr.Number(value=42, label="Seed (Try a different value if the result is unsatisfying)", precision=0)
|
280 |
+
|
281 |
+
sample_steps = gr.Slider(
|
282 |
+
label="Sample Steps",
|
283 |
+
minimum=30,
|
284 |
+
maximum=75,
|
285 |
+
value=75,
|
286 |
+
step=5
|
287 |
+
)
|
288 |
+
|
289 |
+
with gr.Row():
|
290 |
+
submit = gr.Button("Generate", elem_id="generate", variant="primary")
|
291 |
+
|
292 |
+
with gr.Row(variant="panel"):
|
293 |
+
gr.Examples(
|
294 |
+
examples=[
|
295 |
+
os.path.join("examples", img_name) for img_name in sorted(os.listdir("examples"))
|
296 |
+
],
|
297 |
+
inputs=[input_image],
|
298 |
+
label="Examples",
|
299 |
+
examples_per_page=15
|
300 |
+
)
|
301 |
+
|
302 |
+
with gr.Column():
|
303 |
+
|
304 |
+
with gr.Row():
|
305 |
+
|
306 |
+
with gr.Column():
|
307 |
+
mv_show_images = gr.Image(
|
308 |
+
label="Generated Multi-views",
|
309 |
+
type="pil",
|
310 |
+
width=379,
|
311 |
+
interactive=False
|
312 |
+
)
|
313 |
+
|
314 |
+
with gr.Column():
|
315 |
+
output_video = gr.Video(
|
316 |
+
label="video", format="mp4",
|
317 |
+
width=379,
|
318 |
+
autoplay=True,
|
319 |
+
interactive=False
|
320 |
+
)
|
321 |
+
|
322 |
+
with gr.Row():
|
323 |
+
output_model_obj = gr.Model3D(
|
324 |
+
label="Output Model (OBJ Format)",
|
325 |
+
width=768,
|
326 |
+
interactive=False,
|
327 |
+
)
|
328 |
+
gr.Markdown(_LINKS_)
|
329 |
+
gr.Markdown(_CITE_)
|
330 |
+
|
331 |
+
submit.click(fn=check_input_image, inputs=[input_image]).success(
|
332 |
+
fn=preprocess,
|
333 |
+
inputs=[input_image, do_remove_background],
|
334 |
+
outputs=[processed_image],
|
335 |
+
).success(
|
336 |
+
fn=make3d,
|
337 |
+
inputs=[processed_image, sample_steps, sample_seed],
|
338 |
+
outputs=[output_video, output_model_obj, mv_show_images]
|
339 |
+
)
|
340 |
+
|
341 |
+
demo.launch()
|