2727
2828
2929 < title > Image Reconstruction as a Tool for Feature Analysis</ title >
30- < link rel ="icon " type ="image/x-icon " href ="static/images/favicon.ico ">
30+ < link rel ="icon " type ="image/x-icon " href ="static/images/favicon.png ">
3131 < link href ="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro " rel ="stylesheet ">
3232
3333 < link rel ="stylesheet " href ="static/css/bulma.min.css ">
5555 < div class ="column has-text-centered ">
5656 < h1 class ="title is-1 publication-title "> Image Reconstruction as a Tool for Feature Analysis</ h1 >
5757 < div class ="is-size-5 publication-authors ">
58- <!-- Paper authors -->
5958 < span class ="author-block ">
60- < a href =" mailto:allakhverdov@2a2i.org " target =" _blank " > Eduard Allakhverdov</ a >
59+ < strong > Eduard Allakhverdov</ strong > < sup > §† </ sup > ,
6160 </ span >
6261 < span class ="author-block ">
63- < a href =" mailto:d.tarasov@airi.net " target =" _blank " > Dmitrii Tarasov</ a >
62+ < strong > Dmitrii Tarasov</ strong > < sup > § </ sup > ,
6463 </ span >
6564 < span class ="author-block ">
66- < a href =" mailto:goncharova@airi.net " target =" _blank " > Elizaveta Goncharova</ a >
65+ < strong > Elizaveta Goncharova</ strong > < sup > § </ sup > ,
6766 </ span >
6867 < span class ="author-block ">
69- < a href ="mailto:kuznetsov@airi.net " target ="_blank "> Andrey Kuznetsov</ a >
70- </ span >
71- </ div >
72- < div class ="is-size-5 publication-authors ">
73- < span class ="author-block ">
74- AIRI, Moscow, Russia< br >
75- MIPT, Dolgoprudny, Russia
76- </ span >
77- < span class ="author-block ">
78- AIRI, Moscow, Russia
79- </ span >
80- < span class ="author-block ">
81- AIRI, Moscow, Russia
82- </ span >
83- < span class ="author-block ">
84- AIRI, Moscow, Russia
68+ < strong > Andrey Kuznetsov</ strong > < sup > §</ sup > ,
8569 </ span >
70+ < ul >
71+ < sup > §</ sup > AIRI, < sup > †</ sup > MIPT
72+ </ ul >
8673 </ div >
8774
8875 < div class ="column has-text-centered ">
@@ -100,6 +87,7 @@ <h1 class="title is-1 publication-title">Image Reconstruction as a Tool for Feat
10087 </ span >
10188
10289 <!-- ArXiv abstract Link -->
90+ <!-- TODO: add arXiv link -->
10391 < span class ="link-block ">
10492 < a href ="https://arxiv.org/abs/<ARXIV PAPER ID> " target ="_blank "
10593 class ="external-link button is-normal is-rounded is-dark ">
@@ -150,45 +138,60 @@ <h2 class="title is-3">Abstract</h2>
150138 <!-- Базовые результаты: siglip vs siglip2 -->
151139 <!-- Нужно четко подчеркнуть какие различия между модельками -->
152140 <!-- И как это влияет на реконструкцию -->
153- <!-- -->
141+
142+ < div class ="columns is-centered has-text-centered ">
143+ < div class ="column is-four-fifths ">
144+ < h2 class ="title is-3 " style ="white-space: nowrap; "> Reconstruct images from feature space</ h2 >
145+ < div class ="content has-text-justified ">
146+ < div style ="text-align: center; ">
147+ < img src ="./static/images/features_reconstruction.drawio.png " alt ="features_reconstruction " width ="900 ">
148+ < p class ="caption " style ="width: 100%; text-align: center; ">
149+ < b > Figure 1. Image reconstructor training.</ b > For pretrained model we train a reconstructor model that restores the image from the feature space.
150+ </ p >
151+ < img src ="./static/images/reconstruction_metrics.jpg " alt ="reconstruction_metrics " width ="900 ">
152+ < p class ="caption " style ="width: 100%; text-align: center; ">
153+ < b > Figure 2. Reconstruction Metrics.</ b > We show the results of the reconstruction for SigLip and SigLip2 for different image resultions.
154+ </ p >
155+ </ div >
156+ </ div >
157+ < br > < br >
158+ </ div >
159+ </ div >
154160
155161 <!-- (2) Feature-space transformations -->
156162 <!-- Текстовое объяснение -->
157163 <!-- Визуализация фреймворка: обобщил оператор в пр-ве картинок и в пр-ве фичей -->
158- <!-- Сделать видос с визуализацией фреймворка -->
159164 <!-- Примеры работы с RGB -->
160165 <!-- Примеры работы с отключением одного канала (ожелтением) -->
166+ <!-- Примеры Спектра такой м-цы, показать, что только небольшое кол-во каналов меняется -->
161167 <!-- -->
162168
163-
164- <!-- Image carousel -->
165- < section class ="hero is-small ">
166- < div class ="hero-body ">
167- < div class ="container ">
168- < div id ="results-carousel " class ="carousel results-carousel ">
169- < div class ="item ">
170- <!-- Your image here -->
171- < img src ="static/images/v1_vs_v2.png " alt ="Comparison of SigLIP and SigLIP2 reconstructions " />
172- < h2 class ="subtitle has-text-centered ">
173- Comparison of image reconstructions from SigLIP and SigLIP2 feature spaces.
174- </ h2 >
169+ < div class ="columns is-centered has-text-centered ">
170+ < div class ="column is-four-fifths ">
171+ < h2 class ="title is-3 " style ="white-space: nowrap; "> Feature-space transformations</ h2 >
172+ < div class ="content has-text-justified ">
173+ < div style ="text-align: center; ">
174+ < img src ="./static/images/features_reconstruction_manipulation.drawio.png " alt ="features_reconstruction_manipulation " width ="900 ">
175+ < p class ="caption " style ="width: 100%; text-align: center; ">
176+ < b > Figure 3. Feature-space transformations.</ b > We then calculate Q matrix for feature-space manupulation.
177+ </ p >
178+ < img src ="./static/images/rb_swap.png " alt ="rb_swap " width ="900 ">
179+ < p class ="caption " style ="width: 100%; text-align: center; ">
180+ < b > Figure 4. Red-blue channel swap samples.</ b > We then calculate Q matrix for feature-space manupulation.
181+ </ p >
182+ < img src ="./static/images/rb_swap_eigenvalues.png " alt ="rb_swap_eigenvalues " width ="900 ">
183+ < p class ="caption " style ="width: 100%; text-align: center; ">
184+ < b > Figure 5. Eigenvalues for red-blue channel swap matrix.</ b > Majority of eigenvalues are close to 1, which means that the transformation is close to an identity matrix. While the other cluster of eigenvalues are close to -1, which means that for these channels direction is changed to the opposite.
185+ </ p >
175186 </ div >
176- < div class ="item ">
177- <!-- Your image here -->
178- < img src ="static/images/rb_swap.png " alt ="Red-Blue channel swap visualization " />
179- < h2 class ="subtitle has-text-centered ">
180- Visualization of feature space manipulation through red-blue channel swap.
181- </ h2 >
182187 </ div >
188+ < br > < br >
183189 </ div >
184190 </ div >
185- </ div >
186- </ section >
187- <!-- End image carousel -->
188-
189191
190192
191193
194+ <!-- TODO: add citation -->
192195 <!--BibTex citation -->
193196 < section class ="section " id ="BibTeX ">
194197 < div class ="container is-max-desktop content ">
0 commit comments