Spaces:
Running
on
Zero
Running
on
Zero
rizavelioglu
commited on
Commit
·
3050d2d
1
Parent(s):
85dc5d9
add explanations, fix img processing, add another vae, examples
Browse files- app.py +47 -17
- examples/01967_00.jpg +0 -0
- examples/03032_00.jpg +0 -0
- examples/048395_0.jpg +0 -0
- examples/048399_0.jpg +0 -0
- examples/048400_0.jpg +0 -0
- examples/048410_0.jpg +0 -0
- examples/048436_0.jpg +0 -0
- examples/051807_0.jpg +0 -0
- examples/051808_0.jpg +0 -0
- examples/051836_0.jpg +0 -0
- examples/053055_0.jpg +0 -0
- examples/053114_0.jpg +0 -0
- examples/053137_0.jpg +0 -0
- examples/07089_00.jpg +0 -0
- examples/13136_00.jpg +0 -0
- examples/13331_00.jpg +0 -0
- examples/13988_00.jpg +0 -0
- examples/14009_00.jpg +0 -0
- examples/14022_00.jpg +0 -0
- examples/14533_00.jpg +0 -0
app.py
CHANGED
@@ -3,7 +3,7 @@ import torch
|
|
3 |
from diffusers import AutoencoderKL
|
4 |
import torchvision.transforms.v2 as transforms
|
5 |
from torchvision.io import read_image
|
6 |
-
from typing import
|
7 |
import os
|
8 |
from huggingface_hub import login
|
9 |
|
@@ -11,17 +11,27 @@ from huggingface_hub import login
|
|
11 |
hf_token = os.getenv("access_token")
|
12 |
login(token=hf_token)
|
13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
class VAETester:
|
15 |
def __init__(self, device: str = "cuda" if torch.cuda.is_available() else "cpu"):
|
16 |
self.device = device
|
17 |
self.input_transform = transforms.Compose([
|
18 |
-
|
19 |
transforms.Resize((512, 512), antialias=True),
|
20 |
transforms.ToDtype(torch.float32, scale=True),
|
21 |
transforms.Normalize(mean=[0.5], std=[0.5]),
|
22 |
])
|
23 |
self.base_transform = transforms.Compose([
|
24 |
-
|
25 |
transforms.Resize((512, 512), antialias=True),
|
26 |
transforms.ToDtype(torch.float32, scale=True),
|
27 |
])
|
@@ -33,9 +43,10 @@ class VAETester:
|
|
33 |
def _load_all_vaes(self) -> Dict[str, AutoencoderKL]:
|
34 |
"""Load all available VAE models"""
|
35 |
vae_configs = {
|
36 |
-
"
|
37 |
-
"
|
38 |
-
"
|
|
|
39 |
"FLUX.1-dev": ("black-forest-labs/FLUX.1-dev", "vae")
|
40 |
}
|
41 |
|
@@ -79,7 +90,6 @@ class VAETester:
|
|
79 |
results[name] = (diff_img, recon_img, score)
|
80 |
return results
|
81 |
|
82 |
-
|
83 |
# Initialize tester
|
84 |
tester = VAETester()
|
85 |
|
@@ -96,21 +106,31 @@ def test_all_vaes(image_path: str, tolerance: float):
|
|
96 |
|
97 |
for name in tester.vae_models.keys():
|
98 |
diff_img, recon_img, score = results[name]
|
99 |
-
diff_images.append(diff_img)
|
100 |
-
recon_images.append(recon_img)
|
101 |
-
scores.append(f"{name}: {score:.
|
102 |
|
103 |
-
return diff_images, recon_images, scores
|
104 |
|
105 |
except Exception as e:
|
106 |
error_msg = f"Error: {str(e)}"
|
107 |
-
return [None], [None],
|
108 |
|
|
|
109 |
|
110 |
# Gradio interface
|
111 |
-
with gr.Blocks(title="VAE Performance Tester") as demo:
|
112 |
-
gr.Markdown("# VAE
|
113 |
-
gr.Markdown("
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
|
115 |
with gr.Row():
|
116 |
with gr.Column(scale=1):
|
@@ -120,7 +140,8 @@ with gr.Blocks(title="VAE Performance Tester") as demo:
|
|
120 |
maximum=0.5,
|
121 |
value=0.1,
|
122 |
step=0.01,
|
123 |
-
label="Difference Tolerance"
|
|
|
124 |
)
|
125 |
submit_btn = gr.Button("Test All VAEs")
|
126 |
|
@@ -128,7 +149,15 @@ with gr.Blocks(title="VAE Performance Tester") as demo:
|
|
128 |
with gr.Row():
|
129 |
diff_gallery = gr.Gallery(label="Difference Maps", columns=4, height=512)
|
130 |
recon_gallery = gr.Gallery(label="Reconstructed Images", columns=4, height=512)
|
131 |
-
scores_output = gr.Textbox(label="
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
|
133 |
submit_btn.click(
|
134 |
fn=test_all_vaes,
|
@@ -138,3 +167,4 @@ with gr.Blocks(title="VAE Performance Tester") as demo:
|
|
138 |
|
139 |
if __name__ == "__main__":
|
140 |
demo.launch()
|
|
|
|
3 |
from diffusers import AutoencoderKL
|
4 |
import torchvision.transforms.v2 as transforms
|
5 |
from torchvision.io import read_image
|
6 |
+
from typing import Dict
|
7 |
import os
|
8 |
from huggingface_hub import login
|
9 |
|
|
|
11 |
hf_token = os.getenv("access_token")
|
12 |
login(token=hf_token)
|
13 |
|
14 |
+
class PadToSquare:
|
15 |
+
"""Custom transform to pad an image to square dimensions"""
|
16 |
+
def __call__(self, img):
|
17 |
+
_, h, w = img.shape # Get the original dimensions
|
18 |
+
max_side = max(h, w)
|
19 |
+
pad_h = (max_side - h) // 2
|
20 |
+
pad_w = (max_side - w) // 2
|
21 |
+
padding = (pad_w, pad_h, max_side - w - pad_w, max_side - h - pad_h)
|
22 |
+
return transforms.functional.pad(img, padding, padding_mode="edge")
|
23 |
+
|
24 |
class VAETester:
|
25 |
def __init__(self, device: str = "cuda" if torch.cuda.is_available() else "cpu"):
|
26 |
self.device = device
|
27 |
self.input_transform = transforms.Compose([
|
28 |
+
PadToSquare(),
|
29 |
transforms.Resize((512, 512), antialias=True),
|
30 |
transforms.ToDtype(torch.float32, scale=True),
|
31 |
transforms.Normalize(mean=[0.5], std=[0.5]),
|
32 |
])
|
33 |
self.base_transform = transforms.Compose([
|
34 |
+
PadToSquare(),
|
35 |
transforms.Resize((512, 512), antialias=True),
|
36 |
transforms.ToDtype(torch.float32, scale=True),
|
37 |
])
|
|
|
43 |
def _load_all_vaes(self) -> Dict[str, AutoencoderKL]:
|
44 |
"""Load all available VAE models"""
|
45 |
vae_configs = {
|
46 |
+
"stable-diffusion-v1-4": ("CompVis/stable-diffusion-v1-4", "vae"),
|
47 |
+
"sd-vae-ft-mse": ("stabilityai/sd-vae-ft-mse", ""),
|
48 |
+
"sdxl-vae": ("stabilityai/sdxl-vae", ""),
|
49 |
+
"stable-diffusion-3-medium": ("stabilityai/stable-diffusion-3-medium-diffusers", "vae"),
|
50 |
"FLUX.1-dev": ("black-forest-labs/FLUX.1-dev", "vae")
|
51 |
}
|
52 |
|
|
|
90 |
results[name] = (diff_img, recon_img, score)
|
91 |
return results
|
92 |
|
|
|
93 |
# Initialize tester
|
94 |
tester = VAETester()
|
95 |
|
|
|
106 |
|
107 |
for name in tester.vae_models.keys():
|
108 |
diff_img, recon_img, score = results[name]
|
109 |
+
diff_images.append((diff_img, name))
|
110 |
+
recon_images.append((recon_img, name))
|
111 |
+
scores.append(f"{name:<25}: {score:.1f}")
|
112 |
|
113 |
+
return diff_images, recon_images, "\n".join(scores)
|
114 |
|
115 |
except Exception as e:
|
116 |
error_msg = f"Error: {str(e)}"
|
117 |
+
return [None], [None], error_msg
|
118 |
|
119 |
+
examples = [f"examples/{img_filename}" for img_filename in sorted(os.listdir("examples/"))]
|
120 |
|
121 |
# Gradio interface
|
122 |
+
with gr.Blocks(title="VAE Performance Tester", css=".monospace-text {font-family: 'Courier New', Courier, monospace;}") as demo:
|
123 |
+
gr.Markdown("# VAE Comparison Tool")
|
124 |
+
gr.Markdown("""
|
125 |
+
Upload an image or select an example to compare how different VAEs reconstruct it. Here's what happens:
|
126 |
+
1. The image is padded to a square and resized to 512x512 pixels.
|
127 |
+
2. Each VAE encodes the image into a latent space and decodes it back.
|
128 |
+
3. The tool then generates:
|
129 |
+
- **Difference Maps**: Black-and-white images showing where the reconstruction differs from the original (white areas indicate differences above the tolerance threshold).
|
130 |
+
- **Reconstructed Images**: The outputs from each VAE.
|
131 |
+
- **Sum of Differences**: A numerical score for each VAE, measuring the total difference in pixels exceeding the tolerance.
|
132 |
+
Use the tolerance slider to adjust the sensitivity.
|
133 |
+
""")
|
134 |
|
135 |
with gr.Row():
|
136 |
with gr.Column(scale=1):
|
|
|
140 |
maximum=0.5,
|
141 |
value=0.1,
|
142 |
step=0.01,
|
143 |
+
label="Difference Tolerance",
|
144 |
+
info="Low tolerance (e.g., 0.01): Highly sensitive, flags small deviations. High tolerance (e.g., 0.5): Less sensitive, flags only large deviations, showing fewer differences.",
|
145 |
)
|
146 |
submit_btn = gr.Button("Test All VAEs")
|
147 |
|
|
|
149 |
with gr.Row():
|
150 |
diff_gallery = gr.Gallery(label="Difference Maps", columns=4, height=512)
|
151 |
recon_gallery = gr.Gallery(label="Reconstructed Images", columns=4, height=512)
|
152 |
+
scores_output = gr.Textbox(label="Sum of difference (lower is better reconstruction)", lines=5, elem_classes="monospace-text")
|
153 |
+
|
154 |
+
if examples:
|
155 |
+
with gr.Column():
|
156 |
+
example_gallery = gr.Examples(
|
157 |
+
examples=examples,
|
158 |
+
inputs=image_input,
|
159 |
+
label="Example Images"
|
160 |
+
)
|
161 |
|
162 |
submit_btn.click(
|
163 |
fn=test_all_vaes,
|
|
|
167 |
|
168 |
if __name__ == "__main__":
|
169 |
demo.launch()
|
170 |
+
|
examples/01967_00.jpg
ADDED
![]() |
examples/03032_00.jpg
ADDED
![]() |
examples/048395_0.jpg
ADDED
![]() |
examples/048399_0.jpg
ADDED
![]() |
examples/048400_0.jpg
ADDED
![]() |
examples/048410_0.jpg
ADDED
![]() |
examples/048436_0.jpg
ADDED
![]() |
examples/051807_0.jpg
ADDED
![]() |
examples/051808_0.jpg
ADDED
![]() |
examples/051836_0.jpg
ADDED
![]() |
examples/053055_0.jpg
ADDED
![]() |
examples/053114_0.jpg
ADDED
![]() |
examples/053137_0.jpg
ADDED
![]() |
examples/07089_00.jpg
ADDED
![]() |
examples/13136_00.jpg
ADDED
![]() |
examples/13331_00.jpg
ADDED
![]() |
examples/13988_00.jpg
ADDED
![]() |
examples/14009_00.jpg
ADDED
![]() |
examples/14022_00.jpg
ADDED
![]() |
examples/14533_00.jpg
ADDED
![]() |