-
Notifications
You must be signed in to change notification settings - Fork 2
Expand file tree
/
Copy pathindex.html
More file actions
797 lines (754 loc) · 49.1 KB
/
index.html
File metadata and controls
797 lines (754 loc) · 49.1 KB
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
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
<html>
<head>
<!-- Google tag (gtag.js) -->
<!-- <script async src="https://www.googletagmanager.com/gtag/js?id=G-VTKHNTKBM4"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-VTKHNTKBM4');
</script> -->
<meta charset="utf-8" />
<title>GrowSplat</title>
<!-- TODO double check these are updated -->
<!-- Website Metadata -->
<meta content="Building temporal digital twins of plants by combining 3D Gaussian Splatting with a robust spatial alignment pipeline"
name="description" />
<meta content="GrowSplat: Constructing Temporal Digital Twins of Plants with
Gaussian Splats" property="og:title" />
<meta content="Constructing a temporal sequence of plant Gaussian Splats from multi-view camera data, then perform a two-stage spatial registration approach to yield a consistent 4D model of plant development." property="og:description" />
<meta content="https://berkeleyautomation.github.io/GrowSplat/data/preview_card.png" property="og:image" />
<meta content="GrowSplat: Constructing Temporal Digital Twins of Plants with Gaussian Splats" property="twitter:title" />
<meta content="Constructing a temporal sequence of plant Gaussian Splats from multi-view camera data, then perform a two-stage spatial registration approach to yield a consistent 4D model of plant development." property="twitter:description" />
<meta content="https://berkeleyautomation.github.io/GrowSplat/data/preview_card.png" property="twitter:image" />
<meta property="og:type" content="website" />
<meta content="summary_large_image" name="twitter:card" />
<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1" />
<!-- Fonts -->
<link href="https://fonts.googleapis.com" rel="preconnect" />
<link href="https://fonts.gstatic.com" rel="preconnect" crossorigin="anonymous" />
<script src="https://ajax.googleapis.com/ajax/libs/webfont/1.6.26/webfont.js" type="text/javascript"></script>
<script
type="text/javascript">WebFont.load({ google: { families: ["Lato:100,100italic,300,300italic,400,400italic,700,700italic,900,900italic", "Montserrat:100,100italic,200,200italic,300,300italic,400,400italic,500,500italic,600,600italic,700,700italic,800,800italic,900,900italic", "Ubuntu:300,300italic,400,400italic,500,500italic,700,700italic", "Open Sans:300,300italic,400,400italic,600,600italic,700,700italic,800,800italic", "Changa One:400,400italic", "Varela Round:400", "Bungee Shade:regular", "Roboto:300,regular,500", "Bungee Outline:regular"] } });</script>
<!--[if lt IE 9]><script src="https://cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.3/html5shiv.min.js" type="text/javascript"></script><![endif]-->
<!-- JQuery, scripts etc -->
<script src="https://ajax.googleapis.com/ajax/libs/jquery/1/jquery.min.js"></script>
<script src="script.js" type="text/javascript"></script>
<script src="js/carousel_utils.js" type="text/javascript"></script>
<!-- Stylesheets; tabler icons, fonts, ...-->
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/@tabler/icons@latest/iconfont/tabler-icons.min.css">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/4.7.0/css/font-awesome.min.css">
<link href="style.css" rel="stylesheet" type="text/css" />
<link href="data/growsplat.png" rel="shortcut icon" type="image/x-icon" />
</head>
<body>
<div>
<!-- Title, Subtitle, Authors -->
<div>
<h1 class="title">
<span style="text-wrap: nowrap">GrowSplat: Constructing Temporal Digital Twins of Plants with </span> <span style="text-wrap: nowrap"> Gaussian Splats</span>
<!-- <span style="text-wrap: nowrap">Robot Do 🦾</span> -->
</h1>
<!-- <h1 class="subheader">Imitating Articulated Object Manipulation with Monocular 4D Reconstruction</h1> -->
<!-- Author list -->
<div class="base-row">
<!-- Refer to how author wrapping is performed in https://brentyi.github.io/tilted/ -->
<div style="display: flex; flex-wrap: wrap; justify-content: center;">
<div>
<a href="https://simeonoa.github.io/" target="_blank" class="author-text">
Simeon Adebola
</a><sup>1</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://ehehee.github.io/" target="_blank" class="author-text">
Shuangyu Xie
</a><sup>1</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://chungmin99.github.io/" target="_blank" class="author-text">
Chung Min Kim
</a><sup>1</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://kerrj.github.io/" target="_blank" class="author-text">
Justin Kerr
</a><sup>1</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://scholar.google.com/citations?user=t_vRU4wAAAAJ&hl=nl/" target="_blank" class="author-text">
Bart M. van Marrewijk
</a><sup>3</sup>
</div>
<div>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://research.wur.nl/en/persons/mieke-van-vlaardingen/" target="_blank" class="author-text">
Mieke van Vlaardingen
</a><sup>3</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://research.wur.nl/en/persons/tim-van-daalen" target="_blank" class="author-text">
Tim van Daalen
</a><sup>3</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://www.wur.nl/en/persons/robert-van-loo-1.htm" target="_blank" class="author-text">
E.N. van Loo
</a><sup>3</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://scholar.google.com/citations?user=vk2qKkYAAAAJ&hl=en" target="_blank" class="author-text">
Jose Luis Susa Rincon
</a><sup>2</sup>
</div>
<div>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://scholar.google.com.au/citations?user=AvvaaJcAAAAJ&hl=en" target="_blank" class="author-text">
Eugen Solowjow
</a><sup>2</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://scholar.google.com.au/citations?user=gmrJR2cAAAAJ&hl=en" target="_blank" class="author-text">
Rick van de Zedde
</a><sup>3</sup>
<div style="width: 1.25em; display: inline-block"></div>
<a href="https://goldberg.berkeley.edu/" target="_blank" class="author-text">
Ken Goldberg
</a><sup>1</sup>
</div>
</div>
</div>
<div style="text-align: center">
<h1 id="uc-berkeley"><sup>1</sup>UC Berkeley </h1>
<h1 id="uc-berkeley"><sup>2</sup>Siemens Research Lab, Berkeley</h1>
<h1 id="uc-berkeley"><sup>3</sup>Netherlands Plant Eco-phenotyping Centre, Wageningen University
and Research </h1>
<!-- <span class="text-star">*</span>
Denotes Equal Contribution -->
</div>
</div>
<!-- Submission status -->
<div class="title-row">
<!-- <h2 class="subheader">CoRL 2024 (Oral)</h1> -->
<h2 class="subheader">CASE 2025</h1>
</div>
<!-- Paper / code / data URLs -->
<!-- TODO: Update arxiv/code/data link -->
<div class="base-row add-top-padding">
<!-- Paper -->
<a href="https://ieeexplore.ieee.org/abstract/document/11163998" target="_blank" class="link-block">
<figure>
<img src="https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cab99df4998decfbf9e218e_paper-01.png"
alt="paper"
srcset="https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cab99df4998decfbf9e218e_paper-01-p-500.png"
style="max-height: 4em" />
</figure>
<figcaption>
<strong class="link-labels-text">Paper </strong>
</figcaption>
<!-- <figcaption>
<strong class="link-labels-text">Coming Soon </strong>
</figcaption> -->
</a>
<!-- Code -->
<a href="https://berkeleyautomation.github.io/GrowSplat/" target="_blank" class="link-block">
<figure>
<img src="https://uploads-ssl.webflow.com/51e0d73d83d06baa7a00000f/5cae3b53b42ebb3dd4175a82_68747470733a2f2f7777772e69636f6e66696e6465722e636f6d2f646174612f69636f6e732f6f637469636f6e732f313032342f6d61726b2d6769746875622d3235362e706e67.png"
alt="code" style="max-height: 4em" />
</figure>
<figcaption>
<strong class="link-labels-text"></Code ></strong>
</figcaption>
<figcaption>
<strong class="link-labels-text">Coming Soon </strong>
</figcaption>
</a>
<!-- Data -->
<a href="https://huggingface.co/datasets/SimeonOA/GrowSplat-Dataset" target="_blank" class="link-block">
<figure>
<img src="data/database_icon.jpg" alt="data" style="max-height: 4em" />
</figure>
<figcaption>
<strong class="link-labels-text">Data </strong>
</figcaption>
<!-- <figcaption>
<strong class="link-labels-text">Coming Soon </strong>
</figcaption> -->
</a>
</div>
<!-- TL;DR + Teaser video -->
<!-- <div class="section base-row add-top-padding">
<h1 class="tldr">
<b>TL;DR</b>:
GrowSplat uses Gaussian splatting, plant registration and view rendering to create temporal digital twins of plant from industrial scale data.
</h1>
<video id="main-video" autobuffer muted autoplay loop controls playsinline>
<source id="mp4" src="data/supp_video.mp4" type="video/mp4">
</video>
</div> -->
<!-- Abstract -->
<div class="section base-row add-top-padding">
<h1>Abstract</h1>
<p class="paragraph">
Accurate temporal reconstructions of plant growth are essential for plant phenotyping and breeding, yet remain challenging due to complex geometries, occlusions, and non-rigid deformations of plants. We present a novel framework for building temporal digital twins of plants by combining 3D
Gaussian Splatting with a robust sample alignment pipeline. Our method begins by reconstructing Gaussian Splats from multi-view camera data, then leverages a two-stage registration approach: coarse alignment through feature-based matching and Fast Global Registration, followed by fine alignment with Iterative Closest Point. This pipeline yields a consistent 4D model of plant development in discrete time steps. We evaluate the approach on data from the Netherlands Plant Eco-phenotyping Center, demonstrating detailed temporal reconstructions of Sequoia and Quinoa species.
</p>
</div>
<!-- <div class="section base-row add-top-padding">
<h1 class="tldr">
Humans imitate manipulation by watching <b>object</b> motion, not <b>hand</b> motion. RSRD does the same. <br>
This enables imitation from a single video robust to orientation and across the human-robot embodiment gap.
</h1>
<img src="data/rsrd_mainfig.jpg" style="max-width: 100%" />
</div>
<div class="section add-top-padding">
<div class="base-row">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
</div>
</div> -->
<div class="section base-row add-top-padding">
<h1>3D Gaussian Splats for Plants</h1>
<p class="paragraph">
GrowSplat takes in <b>1) </b>single images from 15 static cameras and <b>2)</b> corresponding masks and <b>3)</b> a pointcloud .
By using a custom version of <a href="https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/" class="author-text" target="_blank">3D Gaussian Splatting</a> using <a href="https://docs.nerf.studio/nerfology/methods/splat.html" class="author-text"target="_blank">Splatfacto</a> with the
<a href="https://ubc-vision.github.io/3dgs-mcmc/" class="author-text"target="_blank">Markov Chain Monte Carlo</a> strategy,
we generate 3D reconstructions of plants.:
</p>
<div id="main-results">
<h1 class="tldr"><b>Input Marvin Images</b></h1>
<div class="video-carousel-wrapper">
<div class="carousel-container" id="videoCarousel">
<!-- Add your video elements here -->
<div class="carousel-item" id="nov22_marvin">
<img src="data/input_images/nov22_marvin.png" height="400px" alt="November 22">
</div>
<div class="carousel-item" id="nov27_marvin">
<img src="data/input_images/nov27_marvin.png" height="400px" alt="November 27">
</div>
<div class="carousel-item" id="dec07_marvin">
<img src="data/input_images/dec07_marvin.png" height="400px" alt="December 7">
</div>
<div class="carousel-item" id="dec12_marvin">
<img src="data/input_images/dec12_marvin.png" height="400px" alt="December 12">
</div>
<div class="carousel-item" id="dec17_marvin">
<img src="data/input_images/dec17_marvin.png" height="400px" alt="December 17">
</div>
<div class="carousel-item" id="dec22_marvin">
<img src="data/input_images/dec22_marvin.png" height="400px" alt="December 22">
</div>
<div class="carousel-item" id="dec28_marvin">
<img src="data/input_images/dec28_marvin.png" height="400px" alt="December 28">
</div>
<div class="carousel-item" id="jan02_marvin">
<img src="data/input_images/jan02_marvin.png" height="400px" alt="January 2">
</div>
<div class="carousel-item" id="jan07_marvin">
<img src="data/input_images/jan07_marvin.png" height="400px" alt="January 7">
</div>
<div class="carousel-item" id="jan12_marvin">
<img src="data/input_images/jan12_marvin.png" height="400px" alt="January 12">
</div>
</div>
</div>
<h1 class="tldr"><b>
<img src="data/drag_icon.png" alt="" class="inline-image">Click and move the Quinoa Plants across dates! You can zoom in/out, move up/down, and move left/right. For each date, you can see above the 15 camera images from MaxiMarvin.<img src="data/drag_icon.png" alt="" class="inline-image">
</b></h1>
<div id="iframe-container" class="iframe-container">
<div class="click-and-move-overlay">
<h1 class="tldr">
<b>
<img src="data/drag_icon.png" alt="" class="inline-image">
Click and move me!
<img src="data/drag_icon.png" alt="" class="inline-image">
</b>
</h1>
</div>
<iframe
id="nov22"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_nov_22.viser"
></iframe>
<iframe
id = "nov27"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_nov_27.viser"
></iframe>
<iframe
id = "dec07"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_dec_07.viser"
></iframe>
<iframe
id = "dec12"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_dec_12.viser"
></iframe>
<iframe
id = "dec17"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_dec_17.viser&initDistanceScale=0.6"
></iframe>
<iframe
id="dec22"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_dec_22.viser&initDistanceScale=0.8"
></iframe>
<iframe
id="dec28"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_dec_28.viser&initDistanceScale=0.8"
></iframe>
<iframe
id="jan02"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_jan_07.viser&initDistanceScale=0.9"
></iframe>
<iframe
id="jan07"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_jan_02.viser"
></iframe>
<iframe
id="jan12"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/quinoa_jan_12.viser"
></iframe>
</div>
<div style="position: relative; display: flex;">
<button class="results-slide-arrow" id="results-slide-arrow-prev" onclick="results_slide_left()">
‹
</button>
<div class="results-slide-row" id="results-objs-scroll">
<div data-img-src="data/quinoa/thumbnails/nov22_zoomed.jpg" data-id="nov22-thumb" data-label="November 22" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/nov27_zoomed.jpg" data-id="nov27-thumb" data-label="November 27" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/dec07_zoomed.jpg" data-id="dec07-thumb" data-label="December 07" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/dec12_zoomed.jpg" data-id="dec12-thumb" data-label="December 12" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/dec17_zoomed.jpg" data-id="dec17-thumb" data-label="December 17" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/dec22_zoomed.jpg" data-id="dec22-thumb" data-label="December 22" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/dec28_zoomed.jpg" data-id="dec28-thumb" data-label="December 28" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/jan02_zoomed.jpg" data-id="jan02-thumb" data-label="January 02" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/jan07_zoomed.jpg" data-id="jan07-thumb" data-label="January 07" data-type="image" data-section="quinoa"></div>
<div data-img-src="data/quinoa/thumbnails/jan12_zoomed.jpg" data-id="jan12-thumb" data-label="January 12" data-type="image" data-section="quinoa"></div>
</div>
<button class="results-slide-arrow" id="results-slide-arrow-next" onclick="results_slide_right()">
›
</button>
</div>
<div id="main-results">
<h1 class="tldr"><b>Input Marvin Images</b></h1>
<div class="video-carousel-wrapper">
<div class="carousel-container" id="videoCarousel">
<!-- Add your video elements here -->
<div class="carousel-item" id="feb13_marvin">
<img src="data/input_images/feb13_marvin.png" height="400px" alt="February 13">
</div>
<div class="carousel-item" id="feb16_marvin">
<img src="data/input_images/feb16_marvin.png" height="400px" alt="February 16">
</div>
<div class="carousel-item" id="feb22_marvin">
<img src="data/input_images/feb22_marvin.png" height="400px" alt="February 22">
</div>
<div class="carousel-item" id="feb26_marvin">
<img src="data/input_images/feb26_marvin.png" height="400px" alt="February 26">
</div>
<div class="carousel-item" id="mar01_marvin">
<img src="data/input_images/mar01_marvin.png" height="400px" alt="March 01">
</div>
<div class="carousel-item" id="mar05_marvin">
<img src="data/input_images/mar05_marvin.png" height="400px" alt="March 05">
</div>
<div class="carousel-item" id="mar09_marvin">
<img src="data/input_images/mar09_marvin.png" height="400px" alt="March 09">
</div>
</div>
</div>
<h1 class="tldr"><b>
<img src="data/drag_icon.png" alt="" class="inline-image">Click and move the Sequoia Plants across dates! You can zoom in/out, move up/down, and move left/right. For each date, you can see above the 15 camera images from MaxiMarvin.<img src="data/drag_icon.png" alt="" class="inline-image">
</b></h1>
<div id="iframe-container" class="iframe-container">
<div class="click-and-move-overlay">
<h1 class="tldr">
<b>
<img src="data/drag_icon.png" alt="" class="inline-image">
Click and move me!
<img src="data/drag_icon.png" alt="" class="inline-image">
</b>
</h1>
</div>
<iframe
id="feb13"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_feb_13.viser&initDistanceScale=0.4"
></iframe>
<iframe
id = "feb16"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_feb_16.viser&initDistanceScale=0.4"
></iframe>
<iframe
id = "feb22"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_feb_22.viser&initDistanceScale=0.4"
></iframe>
<iframe
id = "feb26"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_feb_26.viser&initDistanceScale=0.4"
></iframe>
<iframe
id = "mar1"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_mar_01.viser&initDistanceScale=0.43"
></iframe>
<iframe
id = "mar5"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_mar_05.viser&initDistanceScale=0.45"
></iframe>
<iframe
id = "mar9"
class = "iframe"
data-src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/sequoia_mar_09.viser&initDistanceScale=0.48"
></iframe>
</div>
<div style="position: relative; display: flex;">
<button class="results-slide-arrow" id="results-slide-arrow-prev" onclick="results_slide_left()">
‹
</button>
<div class="results-slide-row" id="results-objs-scroll">
<div data-img-src="data/sequoia/thumbnails/feb13_zoomed.jpg" data-id="feb13-thumb" data-label="February 13" data-type="image" data-section="sequoia"></div>
<div data-img-src="data/sequoia/thumbnails/feb16_zoomed.jpg" data-id="feb16-thumb" data-label="February 16" data-type="image" data-section="sequoia"></div>
<div data-img-src="data/sequoia/thumbnails/feb22_zoomed.jpg" data-id="feb22-thumb" data-label="February 22" data-type="image" data-section="sequoia"></div>
<div data-img-src="data/sequoia/thumbnails/feb26_zoomed.jpg" data-id="feb26-thumb" data-label="February 26" data-type="image" data-section="sequoia"></div>
<div data-img-src="data/sequoia/thumbnails/mar1_zoomed.jpg" data-id="mar01-thumb" data-label="March 01" data-type="image" data-section="sequoia"></div>
<div data-img-src="data/sequoia/thumbnails/mar5_zoomed.jpg" data-id="mar05-thumb" data-label="March 05" data-type="image" data-section="sequoia"></div>
<div data-img-src="data/sequoia/thumbnails/mar9_zoomed.jpg" data-id="mar09-thumb" data-label="March 09" data-type="image" data-section="sequoia"></div>
</div>
<button class="results-slide-arrow" id="results-slide-arrow-next" onclick="results_slide_right()">
›
</button>
</div>
</div>
<p class="tldr">These 3D reconstructions are rendered in-browser! If you think that's cool, check out <a href="https://viser.studio/latest/" class="author-text" target="_blank">Viser</a>!</p>
</div>
<div class="section add-top-padding">
<div class="base-row">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
</div>
</div>
<!-- <div class="section base-row add-top-padding">
<h1>Robot Motion Retargeting</h1>
<p class="paragraph">
After recovering 3D part motion, RSRD optimizes grasps and robot motions to reproduce the 4D reconstructions.
</p>
<div id="iframe-container" class="iframe-container">
<div class="click-and-move-overlay">
<h1 class="tldr">
<b>
<img src="data/drag_icon.png" alt="" class="inline-image">
Click and move me!
<img src="data/drag_icon.png" alt="" class="inline-image">
</b>
</h1>
</div>
<iframe class="iframe show"
src="https://berkeleyautomation.github.io/GrowSplat/build/?playbackPath=https://berkeleyautomation.github.io/GrowSplat/recordings/robot_bear.viser&initDistanceScale=1.5&gaussianGroupColorShuffleSeed=1"
></iframe>
</div>
<p class="paragraph">
These can be physically executed on a real robot to produce the demonstrated motion:
</p>
<div style="position: relative; display: flex;">
<button class="results-slide-arrow" id="results-slide-arrow-prev" onclick="vid_slide_left()">
‹
</button>
<div class="video-scroll" id="result-video-scroll">
<div class="video-scroll-card">
<video autoplay muted loop playsinline>
<source src="data/bear_robot_1_trimmed.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-scroll-card">
<video autoplay muted loop playsinline>
<source src="data/nerfgun_robot_1_trimmed.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-scroll-card">
<video autoplay muted loop playsinline>
<source src="data/scissors_white_out.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-scroll-card">
<video autoplay muted loop playsinline>
<source src="data/scissors_red_out.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-scroll-card">
<video autoplay muted loop playsinline>
<source src="data/red_box_out.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-scroll-card">
<video autoplay muted loop playsinline>
<source src="data/sunglasses_out.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<button class="results-slide-arrow" id="results-slide-arrow-next" onclick="vid_slide_right()">
›
</button>
</div>
<p class="paragraph">
RSRD's visual imitation is <b>object-centric</b>, allowing it to adapt to <em>different</em> object orientations with the <em>same</em> demo:
<div class="video-container">
<div class="video-item">
<p><b>0 degrees rotated</b></p>
<video autoplay muted loop playsinline>
<source src="data/bear_robot_2_trimmed.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<p><b>180 degrees rotated</b></p>
<video autoplay muted loop playsinline>
<source src="data/bear_robot_1_trimmed.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<div class="video-container">
<div class="video-item">
<p><b>0 degrees rotated</b></p>
<video autoplay muted loop playsinline>
<source src="data/nerfgun_straight_out.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<p><b>30 degrees rotated</b></p>
<video autoplay muted loop playsinline>
<source src="data/nerfgun_half_turned_out.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<p><b>45 degrees rotated</b></p>
<video autoplay muted loop playsinline>
<source src="data/nerfgun_robot_2_trimmed.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
</p>
</div> -->
<!-- Just for vertical spacing... -->
<!-- <div class="section add-top-padding">
<div class="base-row">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
</div>
</div>
</div>
<div class="section base-row add-top-padding">
<h1>How it works</h1>
<div class="separator"></div>
<p class="tldr"><b>4D Differentiable Part Models</b></p>
<p class="paragraph">
4D-DPM decomposes objects into parts with <a href="https://www.garfield.studio/" class="author-text" target="_blank">GARField</a>, and trains part-centric feature fields
on top of these. Each part is assigned a trainable 6D pose parameter which is optimized with gradient descent.
DINO improves dramatically over photometric tracking as a more robust feature target, and allows reconstructing a broad range of open-world objects.
</p>
<img src="data/4ddpm_fig.jpg" style="max-width: 100%; padding-bottom: 30px" />
<p class="paragraph">
Because 4D-DPM uses gradient descent, any differentiable prior is easily incorporated like temporal smoothness and rigidity.
</p>
<img src="data/4ddpm_ablation.jpg" style="max-width: 90%" />
</div> -->
<div class="section base-row add-top-padding">
<h1>Plant Registration</h1>
<br>
<p class="paragraph">
<ol class="paragraph">
We propose the algorithm framework as follows:
<li>
<b>Point Cloud Initialization and Preprocessing</b>:
Gaussian splats provide a dense point cloud with an associated uncertainty model; however, using the raw data directly
leads to excessive computational overhead and introduces outliers stemming from the uncertainty. To address these
challenges, we propose a point cloud downsampling pipeline for Gaussian splats that applies three principal filtering con-
ditions with Gaussian splats properties to ensure physically valid and well-conditioned data. First, we discard splats with
excessively large or small scales by inspecting their log-scale range, thus preventing skewed modeling. Second, we
compute the scale ratio to remove overly elongated splats
that do not faithfully represent plant geometry. Lastly, we
validate the rotation parameters by checking the norm of the
associated quaternion to reduce numerical errors. After this
filtering process, we estimate surface normals to support the
computation of Fast Point Feature Histograms (FPFH)[31].
These FPFH descriptors capture local geometry and serve as
key features in subsequent alignment steps.
</li><br>
<li>
<b>Global Registration (Coarse Alignment)</b>:
For an initial, coarse alignment of point clouds across different time steps, we adopt a feature-based matching approach inspired by Fast Global Registration (FGR). Specifically, we first compute FPFH for each point in both the reference point
cloud Pref and the temporal point cloud P<sub>tk </sub> . The FPFH
descriptors encode local geometric properties (e.g., curvature,
normal variation) that are invariant to rigid transformations,
facilitating the discovery of putative correspondences be-
tween the two point clouds.We then utilize a RANSAC-
based scheme for robust outlier rejection. In each iteration
of RANSAC, a minimal set of correspondences is randomly
sampled, and an initial rigid transformation is estimated.
The transform is evaluated against the entire set of FPFH
correspondences to identify and reject outliers. We set the
convergence condition to be (3) or the maximum number of
iterations reach the threshold. The resulting transformation
Ttk approximately aligns Ptk to Pref . Once we have a
candidate set of correspondences that survive RANSAC, we
refine them using the optimization procedure proposed in
FGR, which employs a robust objective to handle remaining
outliers more gracefully than pure least-squares approaches.
By combining FPFH-based matching, RANSAC filtering,
and FGR optimization, the final output of this global reg-
istration step is a coarse but sufficiently accurate alignment
of Ptk to Pref . This ensures a robust starting configuration
for subsequent local (fine) registration.
</li><br>
<li>
<b>Fine Registration (Local Alignment)</b>:
Following coarse
alignment, we refine each temporally acquired point cloud
using standard Iterative Closest Point (ICP), which iteratively
refines the rigid transformation by minimizing the Euclidean
distance between matched points. To further improve alignment, we employ Colored ICP, which incorporates photomet-
ric consistency by minimizing color discrepancies between corresponding points. This color term helps disambiguate
similar geometric features, particularly in organic structures
with subtle variations.
</li>
</ol>
</p>
</div>
<div class="section base-row add-top-padding">
<h1>View Rendering</h1>
<p class="paragraph">
We use the registered plant frame and the reconstructed Gaussian splats to render rotating viewing angles. The final output is a temporally consistent sequence of 3D point clouds and video that best captures the growth dynamics of the plant
<br>
<img src="data/rotating-view.png" style="max-width: 90%" />
</p>
</div>
<div class="section base-row add-top-padding">
<h1>Experiments</h1>
<p class="paragraph">
We evaluated GrowSplat with plant species Sequoia and Quinoa that were captured by NPEC Maxi-Marvin system with longest time series with 55 time points over 76 days
<ol class="paragraph">
<li>
<b>Dataset</b>:<br>
<img src="data/dataset.png" style="max-width: 90%" />
</li><br>
<li>
<b>Results</b>:<br>
Top row: registration result for Sequoia. Bottom row: registration result for Quinoa. We use the blue to indicate the point cloud in the earlier date and the green indicate the point cloud in the later date.
<img src="data/example-registration.png" style="max-width: 90%" />
</li><br>
</ol>
</p>
</div>
<!-- Just for vertical spacing... -->
<!-- <div class="section add-top-padding">
<div class="base-row">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
<img src="data/growsplat.png" style="height: 40px; margin-left: 5px; margin-right: 5px; margin-top: 40px; margin-bottom: 40px;">
</div>
</div> -->
<!-- <div class="section base-row add-top-padding">
<h1>Limitations and Failures</h1>
<p class="paragraph">
4D monocular reconstruction is an extremely under-constrained and chal<
<p class="paragraph">
<b>Camera Angle Sensitivity:</b> Since RSRD uses only a single video, it can be sensitive to camera pose during the demonstration, as illustrated in the tracking failure below.
The same scan of the laptop can work or fail depending on the camera angle.
</p>
<div class="video-container">
<div class="video-item">
<video autoplay muted loop playsinline>
<source src="data/laptop_trackfail.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video autoplay muted loop playsinline>
<source src="data/laptop_tracksucc.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<p class="paragraph">
<b>Difficulty with feature-less parts:</b> 4D-DPM can struggle with parts which look similar from multiple angles, or have
not enough visual features for DINO to pick up on. In these videos the tail of the plushie and cable of the charger rotate along their major axis, making robot
execution difficult.
</p>
<div class="video-container">
<div class="video-item">
<video autoplay muted loop playsinline>
<source src="data/garfield_tailrotate.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
<div class="video-item">
<video autoplay muted loop playsinline>
<source src="data/mac_charger_cablerotate.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<p class="paragraph">
<b>Poor segmentation or reconstruction</b>: If the object scan is poor, or the part segmentation is incomplete or severely over-segmented,
4D-DPM can be unstable and lead to catastrophic failure like below.
</p>
<div class="video-container">
<div class="video-item">
<video autoplay muted loop playsinline>
<source src="data/mac_charger_catastrophic.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
<p class="paragraph">
<b>Hand occlusions</b>: hand occlusions can interfere with part motion recovery, such as with the leg of this sculpture
</p>
<div class="video-container">
<div class="video-item">
<video autoplay muted loop playsinline>
<source src="data/hand_occ.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
</div> -->
<div class="section citation" style="margin-top: 50px">
<h1 id="abstract">Citation </h1>
<p class="paragraph"> If you use this work or find it helpful, please consider citing: (bibtex) </p>
<pre id="codecell0">@inproceedings{adebola2025growsplat,
author={Adebola, Simeon and Xie, Shuangyu and Kim, Chung Min and Kerr, Justin and van Marrewijk, Bart M. and van Vlaardingen, Mieke
and van Daalen, Tim and van Loo, E.N. and Susa Rincon, Jose Luis and Solowjow, Eugen and van Zedde, Rick and Goldberg, Ken},
booktitle={2025 IEEE 21st International Conference on Automation Science and Engineering (CASE)},
title={GrowSplat: Constructing Temporal Digital Twins of Plants with Gaussian Splats},
year={2025},
volume={},
number={},
pages={1766-1773},
doi={10.1109/CASE58245.2025.11163998}}
</pre>
</div>
<footer>
<div class="section" style="margin-top: 40px;">
<p class="paragraph">
The website template is adapted from the <a class="author-text" href="https://robot-see-robot-do.github.io/">Robot See Robot Do</a>
project page.
</p>
</div>
</footer>
</div>
</body>