Demake app.py
Browse filesWe wait for now.
app.py
CHANGED
@@ -13,9 +13,9 @@ import numpy as np
|
|
13 |
from networks.gaussian_predictor import GaussianPredictor
|
14 |
from util.vis3d import save_ply
|
15 |
|
16 |
-
|
17 |
def main():
|
18 |
print("[INFO] Starting main function...")
|
|
|
19 |
if torch.cuda.is_available():
|
20 |
device = "cuda:0"
|
21 |
print("[INFO] CUDA is available. Using GPU device.")
|
@@ -23,25 +23,33 @@ def main():
|
|
23 |
device = "cpu"
|
24 |
print("[INFO] CUDA is not available. Using CPU device.")
|
25 |
|
|
|
26 |
print("[INFO] Downloading model configuration...")
|
27 |
-
model_cfg_path = hf_hub_download(repo_id="einsafutdinov/flash3d",
|
|
|
28 |
print("[INFO] Downloading model weights...")
|
29 |
-
model_path = hf_hub_download(repo_id="einsafutdinov/flash3d",
|
|
|
30 |
|
|
|
31 |
print("[INFO] Loading model configuration...")
|
32 |
cfg = OmegaConf.load(model_cfg_path)
|
33 |
|
|
|
34 |
print("[INFO] Initializing GaussianPredictor model...")
|
35 |
model = GaussianPredictor(cfg)
|
36 |
device = torch.device(device)
|
37 |
-
model.to(device)
|
38 |
|
|
|
39 |
print("[INFO] Loading model weights...")
|
40 |
model.load_model(model_path)
|
41 |
|
42 |
-
|
43 |
-
|
|
|
44 |
|
|
|
45 |
def check_input_image(input_image):
|
46 |
print("[DEBUG] Checking input image...")
|
47 |
if input_image is None:
|
@@ -49,27 +57,47 @@ def main():
|
|
49 |
raise gr.Error("No image uploaded!")
|
50 |
print("[INFO] Input image is valid.")
|
51 |
|
52 |
-
|
|
|
53 |
print("[DEBUG] Preprocessing image...")
|
54 |
-
|
|
|
|
|
|
|
|
|
|
|
55 |
image = pad_border_fn(image)
|
56 |
print("[INFO] Image preprocessing complete.")
|
57 |
return image
|
58 |
|
59 |
-
|
60 |
-
|
|
|
|
|
|
|
|
|
61 |
print("[DEBUG] Starting reconstruction and export...")
|
|
|
62 |
image = to_tensor(image).to(device).unsqueeze(0)
|
63 |
-
inputs = {
|
|
|
|
|
|
|
|
|
64 |
print("[INFO] Passing image through the model...")
|
65 |
outputs = model(inputs)
|
|
|
|
|
66 |
print(f"[INFO] Saving output to {ply_out_path}...")
|
67 |
-
save_ply(outputs, ply_out_path, num_gauss=
|
68 |
print("[INFO] Reconstruction and export complete.")
|
|
|
69 |
return ply_out_path
|
70 |
|
|
|
71 |
ply_out_path = f'./mesh.ply'
|
72 |
|
|
|
73 |
css = """
|
74 |
h1 {
|
75 |
text-align: center;
|
@@ -77,15 +105,30 @@ def main():
|
|
77 |
}
|
78 |
"""
|
79 |
|
|
|
80 |
with gr.Blocks(css=css) as demo:
|
81 |
-
gr.Markdown(
|
|
|
|
|
|
|
|
|
82 |
with gr.Row(variant="panel"):
|
83 |
with gr.Column(scale=1):
|
84 |
with gr.Row():
|
85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
with gr.Row():
|
|
|
87 |
submit = gr.Button("Generate", elem_id="generate", variant="primary")
|
|
|
88 |
with gr.Row(variant="panel"):
|
|
|
89 |
gr.Examples(
|
90 |
examples=[
|
91 |
'./demo_examples/bedroom_01.png',
|
@@ -100,29 +143,36 @@ def main():
|
|
100 |
label="Examples",
|
101 |
examples_per_page=20,
|
102 |
)
|
|
|
103 |
with gr.Row():
|
|
|
104 |
processed_image = gr.Image(label="Processed Image", interactive=False)
|
|
|
105 |
with gr.Column(scale=2):
|
106 |
with gr.Row():
|
107 |
with gr.Tab("Reconstruction"):
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
|
|
|
|
112 |
|
|
|
113 |
submit.click(fn=check_input_image, inputs=[input_image]).success(
|
114 |
fn=preprocess,
|
115 |
-
inputs=[input_image
|
116 |
outputs=[processed_image],
|
117 |
).success(
|
118 |
fn=reconstruct_and_export,
|
119 |
-
inputs=[processed_image
|
120 |
outputs=[output_model],
|
121 |
)
|
122 |
|
|
|
123 |
demo.queue(max_size=1)
|
124 |
print("[INFO] Launching Gradio demo...")
|
125 |
-
demo.launch(share=True)
|
126 |
|
127 |
if __name__ == "__main__":
|
128 |
print("[INFO] Running application...")
|
|
|
13 |
from networks.gaussian_predictor import GaussianPredictor
|
14 |
from util.vis3d import save_ply
|
15 |
|
|
|
16 |
def main():
|
17 |
print("[INFO] Starting main function...")
|
18 |
+
# Determine if CUDA (GPU) is available and set the device accordingly
|
19 |
if torch.cuda.is_available():
|
20 |
device = "cuda:0"
|
21 |
print("[INFO] CUDA is available. Using GPU device.")
|
|
|
23 |
device = "cpu"
|
24 |
print("[INFO] CUDA is not available. Using CPU device.")
|
25 |
|
26 |
+
# Download model configuration and weights from Hugging Face Hub
|
27 |
print("[INFO] Downloading model configuration...")
|
28 |
+
model_cfg_path = hf_hub_download(repo_id="einsafutdinov/flash3d",
|
29 |
+
filename="config_re10k_v1.yaml")
|
30 |
print("[INFO] Downloading model weights...")
|
31 |
+
model_path = hf_hub_download(repo_id="einsafutdinov/flash3d",
|
32 |
+
filename="model_re10k_v1.pth")
|
33 |
|
34 |
+
# Load model configuration using OmegaConf
|
35 |
print("[INFO] Loading model configuration...")
|
36 |
cfg = OmegaConf.load(model_cfg_path)
|
37 |
|
38 |
+
# Initialize the GaussianPredictor model with the loaded configuration
|
39 |
print("[INFO] Initializing GaussianPredictor model...")
|
40 |
model = GaussianPredictor(cfg)
|
41 |
device = torch.device(device)
|
42 |
+
model.to(device) # Move the model to the specified device (CPU or GPU)
|
43 |
|
44 |
+
# Load the pre-trained model weights
|
45 |
print("[INFO] Loading model weights...")
|
46 |
model.load_model(model_path)
|
47 |
|
48 |
+
# Define transformation functions for image preprocessing
|
49 |
+
pad_border_fn = TT.Pad((cfg.dataset.pad_border_aug, cfg.dataset.pad_border_aug)) # Padding to augment the image borders
|
50 |
+
to_tensor = TT.ToTensor() # Convert image to tensor
|
51 |
|
52 |
+
# Function to check if an image is uploaded by the user
|
53 |
def check_input_image(input_image):
|
54 |
print("[DEBUG] Checking input image...")
|
55 |
if input_image is None:
|
|
|
57 |
raise gr.Error("No image uploaded!")
|
58 |
print("[INFO] Input image is valid.")
|
59 |
|
60 |
+
# Function to preprocess the input image before passing it to the model
|
61 |
+
def preprocess(image):
|
62 |
print("[DEBUG] Preprocessing image...")
|
63 |
+
# Resize the image to the desired height and width specified in the configuration
|
64 |
+
image = TTF.resize(
|
65 |
+
image, (cfg.dataset.height, cfg.dataset.width),
|
66 |
+
interpolation=TT.InterpolationMode.BICUBIC
|
67 |
+
)
|
68 |
+
# Apply padding to the image
|
69 |
image = pad_border_fn(image)
|
70 |
print("[INFO] Image preprocessing complete.")
|
71 |
return image
|
72 |
|
73 |
+
# Function to reconstruct the 3D model from the input image and export it as a PLY file
|
74 |
+
@spaces.GPU(duration=120) # Decorator to allocate a GPU for this function during execution
|
75 |
+
def reconstruct_and_export(image):
|
76 |
+
"""
|
77 |
+
Passes image through model, outputs reconstruction in form of a dict of tensors.
|
78 |
+
"""
|
79 |
print("[DEBUG] Starting reconstruction and export...")
|
80 |
+
# Convert the preprocessed image to a tensor and move it to the specified device
|
81 |
image = to_tensor(image).to(device).unsqueeze(0)
|
82 |
+
inputs = {
|
83 |
+
("color_aug", 0, 0): image,
|
84 |
+
}
|
85 |
+
|
86 |
+
# Pass the image through the model to get the output
|
87 |
print("[INFO] Passing image through the model...")
|
88 |
outputs = model(inputs)
|
89 |
+
|
90 |
+
# Export the reconstruction to a PLY file
|
91 |
print(f"[INFO] Saving output to {ply_out_path}...")
|
92 |
+
save_ply(outputs, ply_out_path, num_gauss=2)
|
93 |
print("[INFO] Reconstruction and export complete.")
|
94 |
+
|
95 |
return ply_out_path
|
96 |
|
97 |
+
# Path to save the output PLY file
|
98 |
ply_out_path = f'./mesh.ply'
|
99 |
|
100 |
+
# CSS styling for the Gradio interface
|
101 |
css = """
|
102 |
h1 {
|
103 |
text-align: center;
|
|
|
105 |
}
|
106 |
"""
|
107 |
|
108 |
+
# Create the Gradio user interface
|
109 |
with gr.Blocks(css=css) as demo:
|
110 |
+
gr.Markdown(
|
111 |
+
"""
|
112 |
+
# Flash3D
|
113 |
+
"""
|
114 |
+
)
|
115 |
with gr.Row(variant="panel"):
|
116 |
with gr.Column(scale=1):
|
117 |
with gr.Row():
|
118 |
+
# Input image component for the user to upload an image
|
119 |
+
input_image = gr.Image(
|
120 |
+
label="Input Image",
|
121 |
+
image_mode="RGBA",
|
122 |
+
sources="upload",
|
123 |
+
type="pil",
|
124 |
+
elem_id="content_image",
|
125 |
+
)
|
126 |
with gr.Row():
|
127 |
+
# Button to trigger the generation process
|
128 |
submit = gr.Button("Generate", elem_id="generate", variant="primary")
|
129 |
+
|
130 |
with gr.Row(variant="panel"):
|
131 |
+
# Examples panel to provide sample images for users
|
132 |
gr.Examples(
|
133 |
examples=[
|
134 |
'./demo_examples/bedroom_01.png',
|
|
|
143 |
label="Examples",
|
144 |
examples_per_page=20,
|
145 |
)
|
146 |
+
|
147 |
with gr.Row():
|
148 |
+
# Display the preprocessed image (after resizing and padding)
|
149 |
processed_image = gr.Image(label="Processed Image", interactive=False)
|
150 |
+
|
151 |
with gr.Column(scale=2):
|
152 |
with gr.Row():
|
153 |
with gr.Tab("Reconstruction"):
|
154 |
+
# 3D model viewer to display the reconstructed model
|
155 |
+
output_model = gr.Model3D(
|
156 |
+
height=512,
|
157 |
+
label="Output Model",
|
158 |
+
interactive=False
|
159 |
+
)
|
160 |
|
161 |
+
# Define the workflow for the Generate button
|
162 |
submit.click(fn=check_input_image, inputs=[input_image]).success(
|
163 |
fn=preprocess,
|
164 |
+
inputs=[input_image],
|
165 |
outputs=[processed_image],
|
166 |
).success(
|
167 |
fn=reconstruct_and_export,
|
168 |
+
inputs=[processed_image],
|
169 |
outputs=[output_model],
|
170 |
)
|
171 |
|
172 |
+
# Queue the requests to handle them sequentially (to avoid GPU resource conflicts)
|
173 |
demo.queue(max_size=1)
|
174 |
print("[INFO] Launching Gradio demo...")
|
175 |
+
demo.launch(share=True) # Launch the Gradio interface and allow public sharing
|
176 |
|
177 |
if __name__ == "__main__":
|
178 |
print("[INFO] Running application...")
|