6363 < script src ="./js/bulma-carousel.min.js "> </ script >
6464 < script src ="./js/bulma-slider.min.js "> </ script >
6565 < script src ="./js/fontawesome.all.min.js "> </ script >
66- < script src ="./js/pcd-viewer-v3 .js " type ="module "> </ script >
66+ < script src ="./js/pcd-viewer-v4 .js " type ="module "> </ script >
6767 < script type ="module " src ="https://esm.sh/gifuct-js@2.1.2 "> </ script >
6868 < script src ="./js/slider.js " type ="module "> </ script >
6969
@@ -189,7 +189,7 @@ <h1 class="title is-2 publication-title">
189189 < div class ="column is-five-fifths ">
190190 < div class ="">
191191 < h3 class ="title is-4 section-title "> Abstract</ h3 >
192- < div class ="has-text-justified ">
192+ < div class ="has-text-left ">
193193 < p >
194194 We introduce < span class ="methodname "> Rectified Point Flow</ span > , a unified parameterization that formulates pairwise point cloud registration and multi-part shape assembly as a single conditional generative problem.
195195 <!-- <br><br> -->
@@ -276,7 +276,7 @@ <h3 class="title is-4 section-title">Abstract</h3>
276276 < div class ="column is-five-fifths ">
277277 < div class ="">
278278 < h3 class ="title is-4 section-title "> Framework</ h3 >
279- < div class ="has-text-justified mb-2 narrow-content ">
279+ < div class ="has-text-left mb-2 ">
280280 < p >
281281 Rectified Point Flow supports < span class ="emph "> shape assembly</ span > and < span class ="emph "> pairwise registration</ span > tasks in a single framework. Given a set of unposed part point clouds \(\{\bar {X}_i\}_{i\in\Omega}\), it predicts each part's point cloud at the target assembled state \(\{\hat {X}_i{(0)}\}_{i\in\Omega}\). Subsequently, we solve Procrustes problem via SVD between the condition point cloud \(\bar X_i\) and the estimated point cloud \(\hat X_i(0)\) to recover the rigid transformation \(\hat T_i\) for each non-anchored part.
282282 </ p >
@@ -288,7 +288,7 @@ <h3 class="title is-4 section-title">Framework</h3>
288288 src ="images/overview_flow_asm.png "
289289 alt ="ReStyle3D method teaser figure "
290290 class ="publication-teaser-img framed-image "
291- style ="width: 90 %; "
291+ style ="width: 100 %; "
292292 >
293293 </ div >
294294 </ div >
@@ -305,15 +305,15 @@ <h3 class="title is-4 section-title">Framework</h3>
305305 < div class ="column is-five-fifths ">
306306 <!-- Abstract -->
307307 <!-- <h3 class="title is-3">Abstract</h3>
308- <div class="content has-text-justified ">
308+ <div class="content has-text-left ">
309309 <p>
310310 We introduce
311311 </p>
312312 </div> -->
313313
314314 <!-- Method. -->
315315 <!-- <h2 class="title is-3">How it works</h2>
316- <div class="content has-text-justified ">
316+ <div class="content has-text-left ">
317317 <div class="columns is-centered " style="align-items: center;">
318318 <div class="column is-half">
319319 <img
@@ -335,7 +335,7 @@ <h3 class="title is-4 section-title">Framework</h3>
335335 </div>
336336 </div>
337337 </div>
338- <div class="content has-text-justified ">
338+ <div class="content has-text-left ">
339339 <p>
340340 <b
341341 >Two-stage approach of Pipeline.
@@ -348,7 +348,7 @@ <h3 class="title is-4 section-title">Framework</h3>
348348 < div class ="container is-max-desktop has-text-centered is-centered ">
349349 < h2 class ="title is-4 has-text-centered section-title "> Multi-part Shape Assembly</ h2 >
350350
351- < p class ="has-text-justified " style ="width: 90%; padding-bottom: 2rem; margin: 0 auto; ">
351+ < p class ="has-text-left " style ="padding-bottom: 2rem; margin: 0 auto; ">
352352 We evaluate our method on the multi-part shape assembly task, where the goal is to estimate the poses of multiple parts given their unposed point clouds.
353353
354354 < br > < br >
@@ -357,13 +357,13 @@ <h2 class="title is-4 has-text-centered section-title">Multi-part Shape Assembly
357357
358358 <!-- column labels -->
359359 <!-- the full‐width composite figure -->
360- < div class ="columns is-centered has-text-centered ">
360+ < div class ="columns is-centered has-text-centered " style =" margin-top: 1rem; " >
361361 < figure class ="image is-fullwidth ">
362362 < img
363363 id ="comparison "
364364 src ="./images/result_assembly.png "
365365 alt ="Comparison with other methods "
366- style ="width: 90 %; display: block; margin: 0 auto; "
366+ style ="width: 100 %; display: block; margin: 0 auto; "
367367 class ="framed-image "
368368 >
369369 < figcaption >
@@ -390,14 +390,14 @@ <h2 class="title is-4 has-text-centered">Pairwise Registration</h2>
390390 </section> -->
391391
392392 < div class ="section ">
393- < div class ="container is-max-desktop " style ="width: 90%; margin: 0 auto; margin-top: 1rem; ">
393+ < div class ="container is-max-desktop " style ="margin: 0 auto; margin-top: 1rem; ">
394394 < h2 class ="title is-4 has-text-centered section-title "> Linear Interpolation in Noise Space</ h2 >
395- < p class ="has-text-justified " style ="padding-bottom: 2rem; margin: 0 auto; ">
395+ < p class ="has-text-left " style ="padding-bottom: 2rem; margin: 0 auto; ">
396396 We visualize the linear interpolation in the noise space by generating the assembled point cloud from \( Z(s) \), where \( Z(s) \) interpolates linearly between two Gaussian noise vectors \( Z_0 \) and \( Z_1 \). We observe a continuous, semantically meaningful
397397mapping from Gaussian noise to valid assemblies.
398398 </ p >
399399
400- < h2 class ="title is-6 mb-2 has-text-justified "> Part Interchanging</ h2 >
400+ < h2 class ="title is-6 mb-2 has-text-left "> Part Interchanging</ h2 >
401401 < div class ="columns is-centered has-text-centered ">
402402 < div class ="slider-container "
403403 data-gif-url ="images/linear_interpolation_2.gif "
@@ -436,7 +436,7 @@ <h2 class="title is-6 mb-2 has-text-justified">Part Interchanging</h2>
436436 </ div >
437437 </ div >
438438
439- < h2 class ="title is-6 mb-2 has-text-justified "> Structural Changing</ h2 >
439+ < h2 class ="title is-6 mb-2 has-text-left "> Structural Changing</ h2 >
440440 < div class ="columns is-centered has-text-centered ">
441441 < div class ="slider-container "
442442 data-gif-url ="images/linear_interpolation_1.gif "
@@ -484,7 +484,7 @@ <h2 class="title is-4 has-text-centered section-title">Generalization to Unseen
484484
485485 <!-- Visual Effects. -->
486486 < div class ="column is-max-desktop ">
487- < div class ="content has-text-justified ">
487+ < div class ="content has-text-left ">
488488 < h2 class ="title is-6 mb-2 "> Parts from Same Categories</ h2 >
489489 < p style ="text-align: justify; ">
490490 We test the model’s ability to generalize to unseen assemblies composed of parts from two different objects within the same category.
@@ -503,7 +503,7 @@ <h2 class="title is-6 mb-2">Parts from Same Categories</h2>
503503
504504 <!-- Matting. -->
505505 < div class ="column ">
506- < div class ="content has-text-justified ">
506+ < div class ="content has-text-left ">
507507 < h2 class ="title is-6 mb-2 "> Parts from Different Categories</ h2 >
508508 < div class ="columns is-centered ">
509509 < div class ="column content ">
0 commit comments