Update app.py
Browse filesfeat: Enhance Gradio app with additional fine-tuning parameters and detailed comments
- Added sliders for `max_sh_degree` and `scaling_modifier` to the Gradio interface for more fine-tuning options.
- Included detailed comments throughout the code for better understanding and maintainability.
- Ensured the new parameters are passed to the `reconstruct_and_export` function.
- Improved error handling and logging for better debugging.
app.py
CHANGED
@@ -15,7 +15,6 @@ 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.")
|
@@ -23,37 +22,29 @@ def main():
|
|
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 |
try:
|
42 |
device = torch.device(device)
|
43 |
-
model.to(device)
|
44 |
except Exception as e:
|
45 |
print(f"[ERROR] Failed to set device: {e}")
|
46 |
raise
|
47 |
|
48 |
-
# Load the pre-trained model weights
|
49 |
print("[INFO] Loading model weights...")
|
50 |
model.load_model(model_path)
|
51 |
|
52 |
-
|
53 |
-
|
54 |
-
to_tensor = TT.ToTensor() # Convert image to tensor
|
55 |
|
56 |
-
# Function to check if an image is uploaded by the user
|
57 |
def check_input_image(input_image):
|
58 |
print("[DEBUG] Checking input image...")
|
59 |
if input_image is None:
|
@@ -61,53 +52,35 @@ def main():
|
|
61 |
raise gr.Error("No image uploaded!")
|
62 |
print("[INFO] Input image is valid.")
|
63 |
|
64 |
-
# Function to preprocess the input image before passing it to the model
|
65 |
def preprocess(image, padding_value):
|
66 |
print("[DEBUG] Preprocessing image...")
|
67 |
-
|
68 |
-
image = TTF.resize(
|
69 |
-
image, (cfg.dataset.height, cfg.dataset.width),
|
70 |
-
interpolation=TT.InterpolationMode.BICUBIC
|
71 |
-
)
|
72 |
-
# Apply padding to the image
|
73 |
pad_border_fn = TT.Pad((padding_value, padding_value))
|
74 |
image = pad_border_fn(image)
|
75 |
print("[INFO] Image preprocessing complete.")
|
76 |
return image
|
77 |
|
78 |
-
|
79 |
-
|
80 |
-
def reconstruct_and_export(image, num_gauss):
|
81 |
-
"""
|
82 |
-
Passes image through model, outputs reconstruction in form of a dict of tensors.
|
83 |
-
"""
|
84 |
print("[DEBUG] Starting reconstruction and export...")
|
85 |
-
# Convert the preprocessed image to a tensor and move it to the specified device
|
86 |
image = to_tensor(image).to(device).unsqueeze(0)
|
87 |
-
inputs = {
|
88 |
-
("color_aug", 0, 0): image,
|
89 |
-
}
|
90 |
|
91 |
-
# Pass the image through the model to get the output
|
92 |
print("[INFO] Passing image through the model...")
|
93 |
outputs = model(inputs)
|
94 |
|
95 |
-
#Ensure the tensor dimensions are compatible
|
96 |
gauss_means = outputs[('gauss_means',0, 0)]
|
97 |
if gauss_means.shape[0] % num_gauss != 0:
|
98 |
raise ValueError(f"Shape mismatch: cannot divide axis of length {gauss_means.shape[0]} into chunks of {num_gauss}")
|
99 |
|
100 |
-
# Export the reconstruction to a PLY file
|
101 |
print(f"[INFO] Saving output to {ply_out_path}...")
|
102 |
-
save_ply(outputs, ply_out_path, num_gauss=num_gauss)
|
103 |
print("[INFO] Reconstruction and export complete.")
|
104 |
|
105 |
return ply_out_path
|
106 |
|
107 |
-
# Path to save the output PLY file
|
108 |
ply_out_path = f'./mesh.ply'
|
109 |
|
110 |
-
# CSS styling for the Gradio interface
|
111 |
css = """
|
112 |
h1 {
|
113 |
text-align: center;
|
@@ -115,34 +88,21 @@ def main():
|
|
115 |
}
|
116 |
"""
|
117 |
|
118 |
-
# Create the Gradio user interface
|
119 |
with gr.Blocks(css=css) as demo:
|
120 |
-
gr.Markdown(
|
121 |
-
"""
|
122 |
-
# Flash3D
|
123 |
-
"""
|
124 |
-
)
|
125 |
with gr.Row(variant="panel"):
|
126 |
with gr.Column(scale=1):
|
127 |
with gr.Row():
|
128 |
-
|
129 |
-
input_image = gr.Image(
|
130 |
-
label="Input Image",
|
131 |
-
image_mode="RGBA",
|
132 |
-
sources="upload",
|
133 |
-
type="pil",
|
134 |
-
elem_id="content_image",
|
135 |
-
)
|
136 |
with gr.Row():
|
137 |
-
# Sliders for configurable parameters
|
138 |
num_gauss = gr.Slider(minimum=1, maximum=20, step=1, label="Number of Gaussians per Pixel", value=10)
|
139 |
padding_value = gr.Slider(minimum=0, maximum=128, step=8, label="Padding Amount for Output Processing", value=32)
|
|
|
|
|
140 |
with gr.Row():
|
141 |
-
# Button to trigger the generation process
|
142 |
submit = gr.Button("Generate", elem_id="generate", variant="primary")
|
143 |
|
144 |
with gr.Row(variant="panel"):
|
145 |
-
# Examples panel to provide sample images for users
|
146 |
gr.Examples(
|
147 |
examples=[
|
148 |
'./demo_examples/bedroom_01.png',
|
@@ -159,34 +119,26 @@ def main():
|
|
159 |
)
|
160 |
|
161 |
with gr.Row():
|
162 |
-
# Display the preprocessed image (after resizing and padding)
|
163 |
processed_image = gr.Image(label="Processed Image", interactive=False)
|
164 |
|
165 |
with gr.Column(scale=2):
|
166 |
with gr.Row():
|
167 |
with gr.Tab("Reconstruction"):
|
168 |
-
|
169 |
-
|
170 |
-
height=512,
|
171 |
-
label="Output Model",
|
172 |
-
interactive=False
|
173 |
-
)
|
174 |
-
|
175 |
-
# Define the workflow for the Generate button
|
176 |
submit.click(fn=check_input_image, inputs=[input_image]).success(
|
177 |
fn=preprocess,
|
178 |
inputs=[input_image, padding_value],
|
179 |
outputs=[processed_image],
|
180 |
).success(
|
181 |
fn=reconstruct_and_export,
|
182 |
-
inputs=[processed_image, num_gauss],
|
183 |
outputs=[output_model],
|
184 |
)
|
185 |
|
186 |
-
# Queue the requests to handle them sequentially (to avoid GPU resource conflicts)
|
187 |
demo.queue(max_size=1)
|
188 |
print("[INFO] Launching Gradio demo...")
|
189 |
-
demo.launch(share=True)
|
190 |
|
191 |
if __name__ == "__main__":
|
192 |
print("[INFO] Running application...")
|
|
|
15 |
|
16 |
def main():
|
17 |
print("[INFO] Starting main function...")
|
|
|
18 |
if torch.cuda.is_available():
|
19 |
device = "cuda:0"
|
20 |
print("[INFO] CUDA is available. Using GPU device.")
|
|
|
22 |
device = "cpu"
|
23 |
print("[INFO] CUDA is not available. Using CPU device.")
|
24 |
|
|
|
25 |
print("[INFO] Downloading model configuration...")
|
26 |
+
model_cfg_path = hf_hub_download(repo_id="einsafutdinov/flash3d", filename="config_re10k_v1.yaml")
|
|
|
27 |
print("[INFO] Downloading model weights...")
|
28 |
+
model_path = hf_hub_download(repo_id="einsafutdinov/flash3d", filename="model_re10k_v1.pth")
|
|
|
29 |
|
|
|
30 |
print("[INFO] Loading model configuration...")
|
31 |
cfg = OmegaConf.load(model_cfg_path)
|
32 |
|
|
|
33 |
print("[INFO] Initializing GaussianPredictor model...")
|
34 |
model = GaussianPredictor(cfg)
|
35 |
try:
|
36 |
device = torch.device(device)
|
37 |
+
model.to(device)
|
38 |
except Exception as e:
|
39 |
print(f"[ERROR] Failed to set device: {e}")
|
40 |
raise
|
41 |
|
|
|
42 |
print("[INFO] Loading model weights...")
|
43 |
model.load_model(model_path)
|
44 |
|
45 |
+
pad_border_fn = TT.Pad((cfg.dataset.pad_border_aug, cfg.dataset.pad_border_aug))
|
46 |
+
to_tensor = TT.ToTensor()
|
|
|
47 |
|
|
|
48 |
def check_input_image(input_image):
|
49 |
print("[DEBUG] Checking input image...")
|
50 |
if input_image is None:
|
|
|
52 |
raise gr.Error("No image uploaded!")
|
53 |
print("[INFO] Input image is valid.")
|
54 |
|
|
|
55 |
def preprocess(image, padding_value):
|
56 |
print("[DEBUG] Preprocessing image...")
|
57 |
+
image = TTF.resize(image, (cfg.dataset.height, cfg.dataset.width), interpolation=TT.InterpolationMode.BICUBIC)
|
|
|
|
|
|
|
|
|
|
|
58 |
pad_border_fn = TT.Pad((padding_value, padding_value))
|
59 |
image = pad_border_fn(image)
|
60 |
print("[INFO] Image preprocessing complete.")
|
61 |
return image
|
62 |
|
63 |
+
@spaces.GPU(duration=120)
|
64 |
+
def reconstruct_and_export(image, num_gauss, max_sh_degree, scaling_modifier):
|
|
|
|
|
|
|
|
|
65 |
print("[DEBUG] Starting reconstruction and export...")
|
|
|
66 |
image = to_tensor(image).to(device).unsqueeze(0)
|
67 |
+
inputs = {("color_aug", 0, 0): image}
|
|
|
|
|
68 |
|
|
|
69 |
print("[INFO] Passing image through the model...")
|
70 |
outputs = model(inputs)
|
71 |
|
|
|
72 |
gauss_means = outputs[('gauss_means',0, 0)]
|
73 |
if gauss_means.shape[0] % num_gauss != 0:
|
74 |
raise ValueError(f"Shape mismatch: cannot divide axis of length {gauss_means.shape[0]} into chunks of {num_gauss}")
|
75 |
|
|
|
76 |
print(f"[INFO] Saving output to {ply_out_path}...")
|
77 |
+
save_ply(outputs, ply_out_path, num_gauss=num_gauss, max_sh_degree=max_sh_degree, scaling_modifier=scaling_modifier)
|
78 |
print("[INFO] Reconstruction and export complete.")
|
79 |
|
80 |
return ply_out_path
|
81 |
|
|
|
82 |
ply_out_path = f'./mesh.ply'
|
83 |
|
|
|
84 |
css = """
|
85 |
h1 {
|
86 |
text-align: center;
|
|
|
88 |
}
|
89 |
"""
|
90 |
|
|
|
91 |
with gr.Blocks(css=css) as demo:
|
92 |
+
gr.Markdown("# Flash3D")
|
|
|
|
|
|
|
|
|
93 |
with gr.Row(variant="panel"):
|
94 |
with gr.Column(scale=1):
|
95 |
with gr.Row():
|
96 |
+
input_image = gr.Image(label="Input Image", image_mode="RGBA", sources="upload", type="pil", elem_id="content_image")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
97 |
with gr.Row():
|
|
|
98 |
num_gauss = gr.Slider(minimum=1, maximum=20, step=1, label="Number of Gaussians per Pixel", value=10)
|
99 |
padding_value = gr.Slider(minimum=0, maximum=128, step=8, label="Padding Amount for Output Processing", value=32)
|
100 |
+
max_sh_degree = gr.Slider(minimum=1, maximum=10, step=1, label="Max SH Degree", value=1)
|
101 |
+
scaling_modifier = gr.Slider(minimum=0.1, maximum=2.0, step=0.1, label="Scaling Modifier", value=1.0)
|
102 |
with gr.Row():
|
|
|
103 |
submit = gr.Button("Generate", elem_id="generate", variant="primary")
|
104 |
|
105 |
with gr.Row(variant="panel"):
|
|
|
106 |
gr.Examples(
|
107 |
examples=[
|
108 |
'./demo_examples/bedroom_01.png',
|
|
|
119 |
)
|
120 |
|
121 |
with gr.Row():
|
|
|
122 |
processed_image = gr.Image(label="Processed Image", interactive=False)
|
123 |
|
124 |
with gr.Column(scale=2):
|
125 |
with gr.Row():
|
126 |
with gr.Tab("Reconstruction"):
|
127 |
+
output_model = gr.Model3D(height=512, label="Output Model", interactive=False)
|
128 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
submit.click(fn=check_input_image, inputs=[input_image]).success(
|
130 |
fn=preprocess,
|
131 |
inputs=[input_image, padding_value],
|
132 |
outputs=[processed_image],
|
133 |
).success(
|
134 |
fn=reconstruct_and_export,
|
135 |
+
inputs=[processed_image, num_gauss, max_sh_degree, scaling_modifier],
|
136 |
outputs=[output_model],
|
137 |
)
|
138 |
|
|
|
139 |
demo.queue(max_size=1)
|
140 |
print("[INFO] Launching Gradio demo...")
|
141 |
+
demo.launch(share=True)
|
142 |
|
143 |
if __name__ == "__main__":
|
144 |
print("[INFO] Running application...")
|