generated from eliahuhorwitz/Academic-project-page-template
-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
328 lines (283 loc) · 14.8 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!-- Meta tags for social media banners, these should be filled in appropriatly as they are your "business card" -->
<!-- Replace the content tag with appropriate information -->
<meta name="description" content="DESCRIPTION META TAG">
<meta property="og:title" content="PCDM"/>
<meta property="og:description" content="Project Page for PCDM"/>
<meta property="og:url" content="perceptual-consistency-in-dm.github.io"/>
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X630-->
<meta property="og:image" content="static/images/manifold-travel.png" />
<meta property="og:image:width" content="1200"/>
<meta property="og:image:height" content="630"/>
<meta name="twitter:title" content="PCDM">
<meta name="twitter:description" content="Project Page for PCDM">
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X600-->
<meta name="twitter:image" content="static/images/manifold.png">
<meta name="twitter:card" content="Project Page for PCDM">
<!-- Keywords for your paper to be indexed by-->
<meta name="keywords" content="">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>PCDM</title>
<link rel="icon" type="image/x-icon" href="">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro"
rel="stylesheet">
<link rel="stylesheet" href="static/css/bulma.min.css">
<link rel="stylesheet" href="static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="static/css/bulma-slider.min.css">
<link rel="stylesheet" href="static/css/fontawesome.all.min.css">
<link rel="stylesheet"
href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="static/css/index.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://documentcloud.adobe.com/view-sdk/main.js"></script>
<script defer src="static/js/fontawesome.all.min.js"></script>
<script src="static/js/bulma-carousel.min.js"></script>
<script src="static/js/bulma-slider.min.js"></script>
<script src="static/js/index.js"></script>
<script type="text/javascript" async
src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/latest.js?config=TeX-MML-AM_CHTML">
</script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
tex2jax: {inlineMath: [['$','$'], ['\\(','\\)']]}
});
</script>
<script type="text/javascript" async
src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/latest.js?config=TeX-MML-AM_CHTML">
</script>
<script
defer
src="https://cdn.jsdelivr.net/npm/img-comparison-slider@8/dist/index.js"
></script>
<link
rel="stylesheet"
href="https://cdn.jsdelivr.net/npm/img-comparison-slider@8/dist/styles.css"
/>
<style>
.eps_null {
color: rgb(53, 144, 243);
}
.eps_cfg {
color: rgb(231, 111, 81);
}
.cfgpp{
color: rgb(42, 157, 143);
}
hr.solid {
border-top: 2px solid #bbb;
}
table, td, th {
border: 1px solid black;
}
table {
border-collapse: collapse;
width: 100%;
}
td {
text-align: center;
vertical-align: middle;
}
</style>
</head>
<body>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title"><span class="pcdm" style="color: rgb(53, 144, 243);">PCDM</span>: <span style="color: rgb(53, 144, 243);">P</span>erceptual <span style="color: rgb(53, 144, 243);">C</span>onsistency in <span style="color: rgb(53, 144, 243);">D</span>iffusion <span style="color: rgb(53, 144, 243);">M</span>odels for No-Reference Image Quality Assessment</h1>
<!-- <div class="is-size-4 publication-authors">
<span class="author-block">
</div> -->
<div class="is-size-4 publication-authors">
<span class="author-block" style="color: rgb(231, 111, 81);; font-weight: bold;"><em>Under Review - ICLR 2025</em></span>
<!-- <br>Conferance name and year</span> -->
</div>
<!-- PDF Link -->
<span class="link-block">
<a href="static/pdfs/PCDM__PERCEPTUAL_CONSISTENCY_IN_DIFFUSION_MODELS_FOR_NO_REFERENCE_IMAGE_QUALITY_AS__SESSMENT.pdf" target="_blank"
class="external-link button is-normal is-rounded is-dark">
</ARXIV>
<span class="icon">
<i class="fas fa-file-pdf "></i>
</span>
<span>PDF</span>
</a>
</span>
<!-- <span class="link-block">
<a href="https://github.com/perceptual-consistency-in-dm/perceptual-consistency-in-dm.github.io/tree/master/static/pdfs/PCDM__PERCEPTUAL_CONSISTENCY_IN_DIFFUSION_MODELS_FOR_NO_REFERENCE_IMAGE_QUALITY_AS__SESSMENT.pdf" target="_red"
class="external-link button is-normal is-rounded is-dark">
</ARXIV>
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span> -->
<!-- Github link -->
<span class="link-block">
<a href="https://github.com/perceptual-consistency-in-dm/pcdm/tree/main" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code (Coming Soon)</span>
</a>
</span>
</div>
</div>
</div>
</div>
</section>
<!-- Teaser Carousel-->
<section class="hero teaser">
<div class="columns is-centered" style="margin-top:-20px">
<h1 class="is-size-3">
<mark style="background: rgb(255, 255, 102)"> PCDM uses pretrained Latent Diffusion Models (LDM) to do
<!-- We decrease CFG gudiance weight to 0 < $ \lambda $ < 1 by revising DDIM renoising step. -->
<!-- Change $ \hat\epsilon^w_c $ → $ \hat\epsilon_\varnothing $ in CFG renoising -->
</div>
<div class="columns is-centered" style="margin-top:-10px">
<h1 class="is-size-3">
<mark style="background: rgb(255, 255, 102)"> Zero-shot No-Reference Image Quality Assessment (NR-IQA).
<!-- Revise DDIM renoising step -->
<!-- ➜ smaller guidance scale 0 < $ \lambda $ < 1 better sampling & inversion -->
<!-- smooth generative trajectory, better sample quality, and superior inversion capability.</mark></h1> -->
<!-- <mark style="background: rgb(255, 255, 102)"> 👉 smaller guidance scale 0 < $ \lambda $ < 1, -->
<!-- better sample & inversion quality -->
</mark></h1>
</div>
<div class="hero-body">
<div class="container">
<!-- add single image for overview-->
<div class="columns is-centered">
<div class="column is-10">
<img src="static/images/manifold-travel.png" alt="manifold-travel"/>
<!-- add centered text -->
<p style="margin-top:5px; margin-bottom: 10px; font-size:18px;">
An overview of our proposed approach: (a) shows the transition of latent samples across
latent manifolds, highlighting the steps of DDIM and Perceptual Manifold Guidance. (b) depicts the content bias (green) on
the manifold <strong>$ \mathcal{M}_0 \approx \mathcal{D}(\mathcal{Z}_0) $</strong> showing that the guidance term in red (PMG) pushes a data sample
<strong>$ x′_{0|t} \sim \mathcal{D}(z′_{0|t})$</strong> towards the perceptually consistent region (yellow) of the manifold.
</p>
</div>
</div>
<!-- Paper abstract -->
<section class="section hero is-light">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-sixths">
<h2 class="title is-3">Abstract</h2>
<div class="content has-text-justified">
<p style="font-size:18px;">
Despite recent advancements in latent diffusion models that generate highdimensional image data and perform various downstream tasks,
there has been little exploration into perceptual consistency within these models on the task of No-Reference Image Quality Assessment (NR-IQA). In this paper, we hypothesize that latent diffusion models implicitly exhibit perceptually consistent local regions within the data manifold. We leverage this insight to guide on-manifold
sampling using perceptual features and input measurements. Specifically, we
propose <strong>Perceptual Manifold Guidance</strong> (PMG), an algorithm that utilizes pretrained latent diffusion models and perceptual quality metrics to obtain perceptually consistent multi-scale and multi-timestep feature maps from the denoising U-Net. We empirically demonstrate that these hyperfeatures exhibit high
correlation with human perception in IQA tasks. Our method can be applied
to any existing pretrained latent diffusion model and is straightforward to integrate. To the best of our knowledge, this paper is the first work to explore
<strong>Perceptual Consistency in Diffusion Models</strong> (PCDM) and apply it to the NR-IQA problem in a zero-shot setting. Extensive experiments on IQA datasets
show that our method, PCDM, achieves state-of-the-art performance, underscoring the superior zero-shot generalization capabilities of diffusion models for NR-IQA tasks.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- End paper abstract -->
<!-- Method -->
<section class="section hero">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-sixths">
<h2 class="title is-3">Proposed Method</h2>
<div class="content has-text-justified">
<div class="columns is-centered">
<div class="column">
<img src="static/images/algo.png" alt="algorithm" class="is-2" style="width: 700px; height: auto;"/>
<p style="text-align: center;">(a)</p>
</div>
<div class="column">
<img src="static/images/PMG.png" alt="PMG" class="is-2" style="width: 359px; height: auto;"/>
<p style="text-align: center;">(b)</p>
</div>
</div>
<p style=" font-size:18px;">
We introduce the Perceptually Consistent Diffusion Model (PCDM) for No-Reference Image Quality Assessment (NR-IQA),
leveraging the robust representation capabilities of pretrained latent diffusion models (LDMs).
Our method employs Perceptual Manifold Guidance (PMG) to direct the diffusion sampling process toward perceptually consistent regions on the data manifold,
enhancing the model's sensitivity to variations in image quality.
Furthermore, we extract multi-scale and multi-timestep features—termed diffusion hyperfeatures—from the denoising U-Net within the LDM.
These hyperfeatures provide a rich and detailed representation crucial for accurate quality assessment.
As illustrated in Algorithm (a), the overall procedure of our PCDM involves guiding the sampling process and extracting hyperfeatures for quality prediction.
Figure (b) demonstrates the hyperfeature extraction step, highlighting how features are aggregated across different scales and timesteps.
To our knowledge, this is the first work to utilize pretrained LDMs for NR-IQA in a zero-shot manner, offering a novel approach that sets the stage for future research in this domain.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- End Method -->
<!-- NR-IQA -->
<section class="section hero ">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-sxiths">
<h2 class="title is-3">Results</h2>
<h2 class="title is-4">1. Authetic Distortion</h2>
<div class="content has-text-justified" style="margin-top:-15px; margin-bottom: 30px;">
<img src="static/images/authentic.png" alt="authentic"/>
<p style="margin-top:5px; margin-bottom: 10px; font-size:18px;">
Comparison of our proposed PCDM with SOTA NR-IQA methods on PLCC and SRCC Scores for authentic IQA datasets.
</p>
</div>
<h2 class="title is-4"> 2. Synthetic Distortion </h2>
<div class="content has-text-justified" style="margin-top:-15px; margin-bottom: 30px;">
<img src="static/images/synthetic.png" alt="synthetic" style="margin-bottom:20px">
<p style="margin-top:5px; margin-bottom: 30px; font-size: 18px;">
Comparison of our proposed PCDM with SOTA NR-IQA methods on PLCC and SRCC Scores for synthetic IQA datasets.
</p>
</div>
<h2 class="title is-4" style="margin-top:-20px;"> <br> 3. AIGC </h2>
<div class="content has-text-justified">
<img src="static/images/aigc.png" alt="ldis_main"/>
<p style="margin-top:5px; margin-bottom: 10px; font-size: 18px;">
PLCC and SRCC comparison of PCDM on AI Generated Datasets for IQA.
</p>
</div>
</div>
</div>
</div>
</section>
<!--BibTex citation -->
<!-- <section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>BibTex Code Here</code></pre>
</div>
</section> -->
<!--End BibTex citation -->
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This page was built using the <a href="https://github.com/eliahuhorwitz/Academic-project-page-template" target="_blank">Academic Project Page Template</a> which was adopted from the <a href="https://nerfies.github.io" target="_blank">Nerfies</a> project page.
You are free to borrow the of this website, we just ask that you link back to this page in the footer. <br> This website is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/" target="_blank">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
<!-- Statcounter tracking code -->
<!-- You can add a tracker to track page visits by creating an account at statcounter.com -->
<!-- End of Statcounter Code -->
</body>
</html>