-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
537 lines (524 loc) · 57.4 KB
/
Copy pathindex.html
File metadata and controls
537 lines (524 loc) · 57.4 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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>Keio CSG — Research Publications</title>
<link rel="stylesheet" href="assets/css/style.css" />
<!-- Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-R6BS853MB3"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-R6BS853MB3');
</script>
</head>
<body>
<header>
<div class="container">
<div class="header-inner">
<img src="logo/health.webp" alt="Keio CSG Logo" class="lab-logo" />
<div class="header-text">
<p class="lab-abbr">Keio CSG</p>
<h1>Keio Computing and Sensing Group</h1>
<p class="lab-tagline">Research in LiDAR sensing, hardware-efficient AI, and autonomous systems security.</p>
</div>
</div>
</div>
</header>
<div class="container">
<div class="filter-panel">
<div class="highlight-row">
<button class="highlight-toggle" id="toggle-highlight">★ Highlights Only</button>
<span class="highlight-desc">Top-tier venues: CVPR · NDSS · VLSI · ISSCC</span>
</div>
<div class="filter-row">
<span class="filter-label">Topic</span>
<div class="filter-chips" id="filter-tag">
<button class="filter-btn active" data-tag="all">All</button>
<button class="filter-btn" data-tag="LiDAR">LiDAR</button>
<button class="filter-btn" data-tag="Security">Security</button>
<button class="filter-btn" data-tag="Dataset">Dataset</button>
<button class="filter-btn" data-tag="Communication">Communication</button>
<button class="filter-btn" data-tag="Sports">Sports</button>
<button class="filter-btn" data-tag="CIM">CIM</button>
<button class="filter-btn" data-tag="Hardware">Hardware</button>
<button class="filter-btn" data-tag="Edge AI">Edge AI</button>
<button class="filter-btn" data-tag="Transformer">Transformer</button>
<button class="filter-btn" data-tag="Efficiency">Efficiency</button>
</div>
</div>
<div class="filter-row">
<span class="filter-label">Venue</span>
<div class="filter-chips" id="filter-venue">
<button class="filter-btn active" data-venue="all">All</button>
<button class="filter-btn" data-venue="CVPR">CVPR</button>
<button class="filter-btn" data-venue="arXiv">arXiv</button>
<button class="filter-btn" data-venue="IROS">IROS</button>
<button class="filter-btn" data-venue="Sensors J.">Sensors J.</button>
<button class="filter-btn" data-venue="ICRA">ICRA</button>
<button class="filter-btn" data-venue="RA-L">RA-L</button>
<button class="filter-btn" data-venue="NDSS">NDSS</button>
<button class="filter-btn" data-venue="MMSports">MMSports</button>
<button class="filter-btn" data-venue="TVLSI">TVLSI</button>
<button class="filter-btn" data-venue="CVPR Findings">CVPR Findings</button>
<button class="filter-btn" data-venue="ASP-DAC">ASP-DAC</button>
<button class="filter-btn" data-venue="ESSERC">ESSERC</button>
<button class="filter-btn" data-venue="ICCV">ICCV</button>
<button class="filter-btn" data-venue="ICCAD">ICCAD</button>
<button class="filter-btn" data-venue="ISSCC">ISSCC</button>
</div>
</div>
</div>
<main id="project-grid">
<section class="project-section" data-section="Sensor">
<h2 class="section-heading">Sensor</h2>
<div class="section-grid">
<article class="project-card" data-tags="LiDAR, Security, Dataset" data-venue="CVPR" data-highlight="true">
<div class="card-teaser">
<img src="assets/images/ghost-fwl-teaser.svg" alt="Ghost-FWL: A Large-Scale Full-Waveform LiDAR Dataset for Ghost Detection and Removal teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-cvpr">CVPR 2026</span>
<span class="badge badge-status badge-accepted">Accepted</span>
</div>
<h2 class="card-title">Ghost-FWL</h2>
<p class="card-authors"><strong>K. Ikeda</strong>, <strong>R. Hara</strong>, <strong>R. Nagata</strong>, <strong>O. Sako</strong>, Z. Ding, <strong>T. Kado</strong>, <strong>I. Fujioka</strong>, <strong>T. Beppu</strong>, M. Isogawa, <strong>K. Yoshioka</strong></p>
<p class="card-description">The first large-scale annotated mobile full-waveform LiDAR dataset for ghost artifact
detection and removal, with 24,412 frames and 7.5 billion peak-level annotations — 100× more than prior work.
Ghost points arise from multi-path reflections off glass and reflective surfaces,
degrading 3D mapping and object detection in autonomous driving.
Includes FWL-MAE pretraining and transformer-based detection baselines across Ghost, Object, Glass, and Noise categories.</p>
<p class="card-description-ja">自動運転LiDARで発生するガラス等の多重反射による幽霊点群（ゴースト）を検出・除去するための世界初の大規模モバイルFWLデータセット。24,412フレーム・75億ピークのアノテーションを収録し従来最大比100倍の規模を誇る。FWL-MAE事前学習とTransformerベースラインも提供し、自動運転の3Dマッピング・物体検出精度向上に貢献する。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span><span class="tag">Dataset</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2603.28224" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a><a href="https://github.com/Keio-CSG/Ghost-FWL" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a><a href="https://openaccess.thecvf.com/content/CVPR2026/papers/Ikeda_Ghost-FWL_A_Large-Scale_Full-Waveform_LiDAR_Dataset_for_Ghost_Detection_and_CVPR_2026_paper.pdf" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://keio-csg.github.io/Ghost-FWL/" class="link-badge link-project" target="_blank" rel="noopener" aria-label="Project Page"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="none" stroke="white" stroke-width="2" xmlns="http://www.w3.org/2000/svg"><circle cx="12" cy="12" r="10"/><path d="M2 12h20M12 2a15.3 15.3 0 0 1 4 10 15.3 15.3 0 0 1-4 10 15.3 15.3 0 0 1-4-10 15.3 15.3 0 0 1 4-10z"/></svg></span><span class="badge-label">Project Page</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Security" data-venue="arXiv" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/pulsar-net-teaser.svg" alt="Neural Reconstruction of LiDAR Point Clouds under Jamming Attacks via Full-Waveform Representation and Simultaneous Laser Sensing teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-arxiv">arXiv 2026</span>
<span class="badge badge-status badge-preprint">Preprint</span>
</div>
<h2 class="card-title">PULSAR-Net</h2>
<p class="card-authors"><strong>R. Yoshida</strong>, <strong>T. Sato</strong>, <strong>W. Zhang</strong>, <strong>Y. Hayakawa</strong>, <strong>S. Nagai</strong>, T. Kado, T. Beppu, I. Fujioka, Y. Zhong, <strong>K. Yoshioka</strong></p>
<p class="card-description">PULSAR-Net, the first effective defense against LiDAR jamming attacks that blind sensors
by flooding them with high-frequency laser pulses.
By leveraging intermediate full-waveform data (normally discarded after peak detection)
and simultaneous multi-laser sensing patterns, a 3D U-Net with axial spatial-temporal attention
segments attack pulses from legitimate reflections directly in the waveform domain.
Trained exclusively on synthetic data, PULSAR-Net achieves 92% and 73% point-cloud reconstruction
rates for vehicles erased by attacks in real-world static and driving scenarios respectively.</p>
<p class="card-description-ja">高頻度レーザーパルスでLiDARを完全に盲目にするジャミング攻撃に対する世界初の有効な防御手法PULSAR-Netを提案。通常廃棄される中間フルウェーブフォームデータと同時マルチレーザーセンシングの空間時間相関を活用し、軸方向アテンション付きU-Netで攻撃パルスを正当反射から分離する。合成データのみで訓練しながら、実世界の静止・走行シナリオで92%・73%の点群再構成率を達成した。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2604.00371" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Security" data-venue="IROS" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/mirrordrift-teaser.svg" alt="MirrorDrift: Actuated Mirror-Based Attacks on LiDAR SLAM teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-iros">IROS 2026</span>
<span class="badge badge-status badge-accepted">Accepted</span>
</div>
<h2 class="card-title">MirrorDrift</h2>
<p class="card-authors"><strong>R. Nagata</strong>, K. Koide, <strong>K. Ikeda</strong>, <strong>O. Sako</strong>, <strong>S. Horie</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">MirrorDrift introduces a novel physical attack on LiDAR SLAM using an actuated mirror
to reflect and redirect laser pulses, injecting controlled ghost reflections into the sensor.
By precisely controlling the mirror orientation, an adversary can cause persistent
localization drift without direct line-of-sight to the target vehicle.</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2603.11364" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Localization, Construction" data-venue="Sensors J." data-highlight="false">
<div class="card-teaser">
<img src="assets/images/lidar-beacon-teaser.svg" alt="LiDAR Beacon: A Long-Range LiDAR Fiducial Marker for GPS-Free Autonomous Localization of Construction Machinery teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-sensorsj">Sensors J. 2026</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">LiDAR Beacon</h2>
<p class="card-authors"><strong>O. Sako</strong>, K. Koide, <strong>K. Yoshioka</strong></p>
<p class="card-description">A LiDAR fiducial marker achieving 309m recognition distance — approximately 19× longer than prior work —
by encoding IDs into temporal reflection patterns using a single laser return on a PDLC film.
Integrated into an autonomous localization system enabling GPS-free, sub-meter accurate positioning
of construction machinery in GPS-denied environments such as tunnels and open terrain.</p>
<p class="card-description-ja">LiDARの1点反射だけでIDを識別できる新方式の基準マーカー「LiDAR Beacon」を開発。PDLCフィルムの透明/不透明切替による時間的反射パターンを利用し、従来比約19倍となる世界最長309mの識別距離を実現した。GPS・トータルステーション不要の自律測位システムを構築し、砂地環境での長距離走行後も平均誤差0.14m以内の高精度を達成している。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Localization</span><span class="tag">Construction</span></div>
<div class="card-links"><a href="https://ieeexplore.ieee.org/document/11410042/" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://www.keio.ac.jp/ja/press-release/20260324-1/" class="link-badge link-press" target="_blank" rel="noopener" aria-label="Press"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M4 22a2 2 0 0 1-2-2V4a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2v3h2v13a2 2 0 0 1-2 2H4zm14-2h2V9h-2v11zM4 4v16h12V4H4zm2 4h8v2H6V8zm0 4h8v2H6v-2zm0 4h5v2H6v-2z"/></svg></span><span class="badge-label">Press</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Security" data-venue="IROS" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/d-slamspoof-teaser.svg" alt="D-SLAMSpoof: An Environment-Agnostic LiDAR Spoofing Attack using Dynamic Point Cloud Injection teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-iros">IROS 2026</span>
<span class="badge badge-status badge-accepted">Accepted</span>
</div>
<h2 class="card-title">D-SLAMSpoof</h2>
<p class="card-authors"><strong>R. Nagata</strong>, K. Koide, <strong>K. Ikeda</strong>, <strong>O. Sako</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">An environment-agnostic LiDAR spoofing attack that injects dynamically generated
fake point clouds to manipulate SLAM-based localization systems.
Unlike prior spoofing attacks that rely on environment-specific setup,
D-SLAMSpoof adapts point cloud injection in real time, enabling effective attacks
across diverse environments without prior scene knowledge.</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2603.11365" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Security" data-venue="ICRA" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/slamspoof-teaser.svg" alt="SLAMSpoof: Practical LiDAR Spoofing Attacks on Localization Systems Guided by Scan Matching Vulnerability Analysis teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-icra">ICRA 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">SLAMSpoof</h2>
<p class="card-authors"><strong>R. Nagata</strong>, K. Koide, <strong>Y. Hayakawa</strong>, <strong>R. Suzuki</strong>, <strong>K. Ikeda</strong>, <strong>O. Sako</strong>, Q. Chen, T. Sato, <strong>K. Yoshioka</strong></p>
<p class="card-description">Practical LiDAR spoofing attacks targeting SLAM-based localization systems,
guided by scan matching vulnerability analysis.
Demonstrates that adversaries can manipulate vehicle localization by injecting
carefully crafted ghost points into LiDAR scans.</p>
<p class="card-description-ja">スキャンマッチングの脆弱性解析に基づく、SLAM型自己位置推定システムへの実践的LiDARスプーフィング攻撃を提案。慎重に設計した偽造点群の注入により、車両の自己位置認識を攻撃者の意図通りに操作できることを実証した。自動運転システムの安全設計に重要な知見を与える。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2502.13641" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a><a href="https://github.com/Keio-CSG/slamspoof" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Communication" data-venue="RA-L" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/optical-lidar-comm-teaser.svg" alt="Optical LiDAR Communication: Repurposing Existing LiDAR Sensors for Infrastructure-to-Vehicle Communication teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-ral">RA-L 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">Optical LiDAR Comm</h2>
<p class="card-authors"><strong>K. Ikeda</strong>, <strong>Y. Hayakawa</strong>, <strong>R. Suzuki</strong>, S. Nagai, <strong>O. Sako</strong>, <strong>R. Nagata</strong>, <strong>R. Yoshida</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">Enables infrastructure-to-vehicle messaging by repurposing existing LiDAR sensors for optical communication,
eliminating the need for dedicated communication hardware.
Demonstrates practical V2I data transfer using commodity automotive LiDAR units.</p>
<p class="card-description-ja">既存の車載LiDARセンサを光通信チャネルとして転用し、インフラ・車両間通信（I2V）を実現する手法を提案。専用通信ハードウェアを不要とし、市販LiDARユニットのみで実用的なデータ転送を実証した。スマートインフラ連携の低コスト実装に向けた新たなアプローチを示す。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Communication</span></div>
<div class="card-links"><a href="https://sites.google.com/keio.jp/keio-csg/projects" class="link-badge link-project" target="_blank" rel="noopener" aria-label="Project Page"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="none" stroke="white" stroke-width="2" xmlns="http://www.w3.org/2000/svg"><circle cx="12" cy="12" r="10"/><path d="M2 12h20M12 2a15.3 15.3 0 0 1 4 10 15.3 15.3 0 0 1-4 10 15.3 15.3 0 0 1-4-10 15.3 15.3 0 0 1 4-10z"/></svg></span><span class="badge-label">Project Page</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Security" data-venue="NDSS" data-highlight="true">
<div class="card-teaser">
<img src="assets/images/lidar-spoof-highspeed-teaser.svg" alt="On the Realism of LiDAR Spoofing Attacks against Autonomous Driving Vehicle at High Speed and Long Distance teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-ndss">NDSS 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">LiDAR Spoofing (High Speed)</h2>
<p class="card-authors">T. Sato, <strong>R. Suzuki</strong>, <strong>Y. Hayakawa</strong>, <strong>K. Ikeda</strong>, <strong>O. Sako</strong>, <strong>R. Nagata</strong>, Q. Chen, <strong>K. Yoshioka</strong></p>
<p class="card-description">Examines the practical feasibility of LiDAR spoofing attacks on autonomous driving vehicles
at realistic high speeds and long distances.
Reveals that spoofing remains effective under conditions previously assumed to be safe,
challenging existing security assumptions.</p>
<p class="card-description-ja">高速走行・長距離条件下での自動運転車両に対するLiDARスプーフィング攻撃の現実的な有効性を検証した研究。これまで安全と想定されていた条件下でも攻撃が成立することを明らかにし、既存のセキュリティ前提を覆す実験的証拠を提示する。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span></div>
<div class="card-links"><a href="https://sites.google.com/keio.jp/keio-csg/projects/AttackonDrivingVehicle" class="link-badge link-project" target="_blank" rel="noopener" aria-label="Project Page"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="none" stroke="white" stroke-width="2" xmlns="http://www.w3.org/2000/svg"><circle cx="12" cy="12" r="10"/><path d="M2 12h20M12 2a15.3 15.3 0 0 1 4 10 15.3 15.3 0 0 1-4 10 15.3 15.3 0 0 1-4-10 15.3 15.3 0 0 1 4-10z"/></svg></span><span class="badge-label">Project Page</span></a><a href="https://www.ndss-symposium.org/wp-content/uploads/2025-628-paper.pdf" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://www.jst.go.jp/pr/announce/20250225-2/pdf/20250225-2.pdf" class="link-badge link-press" target="_blank" rel="noopener" aria-label="Press"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M4 22a2 2 0 0 1-2-2V4a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2v3h2v13a2 2 0 0 1-2 2H4zm14-2h2V9h-2v11zM4 4v16h12V4H4zm2 4h8v2H6V8zm0 4h8v2H6v-2zm0 4h5v2H6v-2z"/></svg></span><span class="badge-label">Press</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Sports" data-venue="MMSports" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/basket-lidar-teaser.svg" alt="Basket LiDAR: 3D Basketball Player Tracking via LiDAR Sensing teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-mmsports">MMSports 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">Basket LiDAR</h2>
<p class="card-authors"><strong>R. Hayashi</strong>, <strong>K. Torimi</strong>, <strong>R. Nagata</strong>, <strong>K. Ikeda</strong>, <strong>O. Sako</strong>, T. Nakamura, M. Tani, Y. Aoki, <strong>K. Yoshioka</strong></p>
<p class="card-description">3D basketball player tracking using LiDAR sensing for sports analytics.
Enables precise real-time localization of players and ball during games,
providing rich spatial data for performance analysis and coaching insights.</p>
<p class="card-description-ja">LiDARセンシングを用いたバスケットボール選手の3D追跡システム。試合中の選手とボールをリアルタイムで精密に3次元位置推定し、スポーツアナリティクスのための高密度空間データを取得する。コーチングや戦術分析への実用的活用を見据えたシステム構築を行った。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Sports</span></div>
<div class="card-links"><a href="https://sites.google.com/keio.jp/keio-csg/projects/basket-lidar" class="link-badge link-project" target="_blank" rel="noopener" aria-label="Project Page"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="none" stroke="white" stroke-width="2" xmlns="http://www.w3.org/2000/svg"><circle cx="12" cy="12" r="10"/><path d="M2 12h20M12 2a15.3 15.3 0 0 1 4 10 15.3 15.3 0 0 1-4 10 15.3 15.3 0 0 1-4-10 15.3 15.3 0 0 1 4-10z"/></svg></span><span class="badge-label">Project Page</span></a><a href="https://github.com/Keio-CSG/BasketLiDAR" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a></div>
</div>
</article>
<article class="project-card" data-tags="LiDAR, Security" data-venue="NDSS" data-highlight="true">
<div class="card-teaser">
<img src="assets/images/lidar-newgen-teaser.svg" alt="LiDAR Spoofing Meets the New-Gen: Capability Improvements, Broken Assumptions, and New Attack Strategies teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-ndss">NDSS 2024</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">LiDAR New-Gen Spoofing</h2>
<p class="card-authors">T. Sato, <strong>Y. Hayakawa</strong>, <strong>R. Suzuki</strong>, Y. Shiiki, <strong>K. Yoshioka</strong>, Q. Chen</p>
<p class="card-description">Reveals that next-generation LiDAR sensors designed for better range and resolution
inadvertently open new spoofing attack surfaces.
Demonstrates improved attack capabilities that invalidate prior safety assumptions
and introduces novel strategies that bypass state-of-the-art countermeasures,
exposing critical security implications for the evolving autonomous vehicle ecosystem.</p>
<p class="card-description-ja">次世代LiDARセンサに対するスプーフィング攻撃の能力向上と新たな攻撃戦略を明らかにした研究。従来の安全仮定が成立しなくなることを示し、既存の対策を無効化する手法を複数提案している。次世代センサのセキュリティ設計指針を与える重要な知見を含む。</p>
<div class="card-tags"><span class="tag">LiDAR</span><span class="tag">Security</span></div>
<div class="card-links"><a href="https://sites.google.com/keio.jp/keio-csg/projects/new-gen-lidar-sec" class="link-badge link-project" target="_blank" rel="noopener" aria-label="Project Page"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="none" stroke="white" stroke-width="2" xmlns="http://www.w3.org/2000/svg"><circle cx="12" cy="12" r="10"/><path d="M2 12h20M12 2a15.3 15.3 0 0 1 4 10 15.3 15.3 0 0 1-4 10 15.3 15.3 0 0 1-4-10 15.3 15.3 0 0 1 4-10z"/></svg></span><span class="badge-label">Project Page</span></a><a href="https://www.ndss-symposium.org/ndss-paper/lidar-spoofing-meets-the-new-gen-capability-improvements-broken-assumptions-and-new-attack-strategies/" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://www.jst.go.jp/pr/announce/20240213-3/index.html" class="link-badge link-press" target="_blank" rel="noopener" aria-label="Press"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M4 22a2 2 0 0 1-2-2V4a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2v3h2v13a2 2 0 0 1-2 2H4zm14-2h2V9h-2v11zM4 4v16h12V4H4zm2 4h8v2H6V8zm0 4h8v2H6v-2zm0 4h5v2H6v-2z"/></svg></span><span class="badge-label">Press</span></a></div>
</div>
</article>
</div>
</section>
<section class="project-section" data-section="Circuit">
<h2 class="section-heading">Circuit</h2>
<div class="section-grid">
<article class="project-card" data-tags="CIM, Hardware" data-venue="TVLSI" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/mcra-teaser.svg" alt="MCRA: Multicolumn Residue Accumulation Analog Compute-in-Memory Architecture With Time-Domain M-Input ΣΔ ADC teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-tvlsi">TVLSI 2026</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">MCRA</h2>
<p class="card-authors"><strong>X. Li</strong>, <strong>W. Zhang</strong>, <strong>S. Ando</strong>, Z. Wang, J. Lin, <strong>K. Yoshioka</strong></p>
<p class="card-description">A multicolumn residue accumulation analog compute-in-memory architecture
using a time-domain M-input ΣΔ ADC to achieve high-precision analog computation.
Residues from multiple columns are accumulated sequentially to boost ADC effective resolution
without proportional area cost, enabling energy-efficient on-chip inference for edge AI.</p>
<p class="card-description-ja">時間領域M入力ΣΔ ADCを用いたマルチカラム残差累積型アナログCIMアーキテクチャ。複数列の残差を逐次累積することでADC精度を効率的に向上させ、高精度なアナログ計算とエネルギー効率の両立を実現。エッジAIデバイス向けに最適化された次世代CIM設計手法を提案する。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span></div>
<div class="card-links"><a href="https://ieeexplore.ieee.org/document/11421079" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a></div>
</div>
</article>
<article class="project-card" data-tags="Hardware, Edge AI, Transformer" data-venue="CVPR Findings" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/d4c-teaser.svg" alt="D4C: Data-Free Quantization for Contrastive Language-Image Pre-training Models teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-cvprfindings">CVPR Findings 2026</span>
<span class="badge badge-status badge-accepted">Accepted</span>
</div>
<h2 class="card-title">D4C</h2>
<p class="card-authors"><strong>W. Zhang</strong>, Y. Zhong, Z. Ding, <strong>X. Li</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">Addresses data-free quantization of CLIP vision-language models by tackling the core challenge
of insufficient semantic diversity in synthesized calibration samples.
D4C combines prompt-guided semantic alignment, structural contrastive generation for compositional diversity,
and perturbation-aware enhancement — enabling effective W4A4 CLIP compression without original training data.</p>
<p class="card-description-ja">元データを使わずにCLIP（視覚・言語基盤モデル）を量子化するD4Cフレームワーク。合成サンプルの意味多様性不足という課題に対し、プロンプト誘導型セマンティックアライメント・構造的コントラスト生成・摂動ロバスト化の3要素で解決する。W4A4でのCLIP圧縮をはじめてデータなしで実現した。</p>
<div class="card-tags"><span class="tag">Hardware</span><span class="tag">Edge AI</span><span class="tag">Transformer</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2511.15411" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a></div>
</div>
</article>
<article class="project-card" data-tags="CIM, Hardware, Edge AI" data-venue="ASP-DAC" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/bitrom-teaser.svg" alt="BitROM: Weight Reload-Free CiROM Architecture Towards Billion-Parameter 1.58-bit LLM Inference teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-aspdac">ASP-DAC 2026</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">BitROM</h2>
<p class="card-authors"><strong>W. Zhang</strong>, <strong>X. Li</strong>, <strong>S. Ando</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">A weight reload-free Compute-in-Read-Only-Memory (CiROM) architecture for billion-parameter LLM inference
using 1.58-bit (ternary) weights.
Integrates bidirectional ROM arrays, tri-mode local accumulators, and eDRAM KV-cache management,
reducing external DRAM access by 43.6% during decoding.
Achieves 20.8 TOPS/W and 10× area efficiency improvement over prior digital CiROM designs.</p>
<p class="card-description-ja">1.58ビット（三値）重みを用いた10億パラメータ規模LLM推論をウェイトリロードなしで実現するCiROMアーキテクチャ。双方向ROMアレイ・三値局所アキュムレータ・KVキャッシュ用eDRAMを統合し、外部DRAMアクセスを43.6%削減。20.8 TOPS/Wの高効率と先行デジタルCiROM比10倍の面積効率を達成する。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span><span class="tag">Edge AI</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2509.08542" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a><a href="https://github.com/Wenlun-Zhang/BitROM" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a></div>
</div>
</article>
<article class="project-card" data-tags="CIM, Hardware, Edge AI" data-venue="ESSERC" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/saliency-cim-teaser.svg" alt="A 4541 TOPS/W Saliency-Aware Analog Computing In-Memory Macro with Charge-Domain Saliency Detector teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-esserc">ESSERC 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">Saliency-Aware CIM</h2>
<p class="card-authors"><strong>S. Ando</strong>, <strong>S. Miyagi</strong>, <strong>W. Zhang</strong>, <strong>Y.-C. Chen</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">A 4541 TOPS/W analog compute-in-memory macro featuring a charge-domain saliency detector
that dynamically skips computation on non-salient regions.
Achieves state-of-the-art energy efficiency for vision inference at the edge.</p>
<p class="card-description-ja">顕著性検出器を内蔵した4541 TOPS/Wのアナログコンピュートインメモリマクロ。電荷ドメインで非顕著領域の演算を動的にスキップし、視覚推論における圧倒的なエネルギー効率を実現する。エッジAIデバイスの次世代設計に向けた画期的なアーキテクチャを提案する。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span><span class="tag">Edge AI</span></div>
<div class="card-links"></div>
</div>
</article>
<article class="project-card" data-tags="CIM, Hardware" data-venue="TVLSI" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/asim-teaser.svg" alt="ASiM: Modeling and Analyzing Inference Accuracy of SRAM-Based Analog CiM Circuits teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-tvlsi">TVLSI 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">ASiM</h2>
<p class="card-authors"><strong>W. Zhang</strong>, <strong>S. Ando</strong>, <strong>Y.C. Chen</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">A simulation framework for modeling and analyzing how analog noise impacts inference accuracy
in SRAM-based compute-in-memory (CiM) circuits.
Reveals that even 1 LSB of analog noise can significantly impair performance on complex tasks like ImageNet,
and evaluates mitigation strategies — hybrid analog-digital execution and majority voting —
to recover accuracy while preserving CiM energy efficiency benefits.</p>
<p class="card-description-ja">SRAMベースのアナログCiM回路における推論精度影響をモデル化・解析するシミュレーションフレームワーク。アナログノイズが1LSBでもImageNet等の複雑タスクで精度を大幅低下させることを示し、ハイブリッドアナログ-デジタル実行や多数決投票による軽減策を評価。省エネ効果を保ちながら精度を確保する設計指針を提供する。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span></div>
<div class="card-links"><a href="https://doi.org/10.1109/tvlsi.2025.3605286" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://github.com/Keio-CSG/ASiM" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a></div>
</div>
</article>
<article class="project-card" data-tags="Hardware, Edge AI, Transformer" data-venue="ICCV" data-highlight="true">
<div class="card-teaser">
<img src="assets/images/ahcptq-teaser.svg" alt="AHCPTQ: Accurate and Hardware-Compatible Post-Training Quantization for Segment Anything Model teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-iccv">ICCV 2025</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">AHCPTQ</h2>
<p class="card-authors"><strong>W. Zhang</strong>, Y. Zhong, <strong>S. Ando</strong>, <strong>K. Yoshioka</strong></p>
<p class="card-description">Enables the first functional 4-bit post-training quantization of the Segment Anything Model (SAM)
via two hardware co-designed innovations: Hybrid Log-Uniform Quantization (HLUQ) for heavy-tailed
post-GELU activations, and Channel-Aware Grouping (CAG) cutting on-chip register overhead by 99.7%.
FPGA deployment achieves 7.89× speedup and 8.64× energy efficiency over float at W4A4 with 36.6% mAP.</p>
<p class="card-description-ja">Segment Anything Model（SAM）の世界初4ビット量子化PTQを実現するハードウェア協調設計手法。post-GELUの重尾分布に対応するHybrid Log-Uniform量子化と、チャネルグルーピングによるオンチップレジスタ削減（−99.7%）を組み合わせる。FPGA上でW4A4動作時に浮動小数点比7.89倍の高速化・8.64倍の省エネを達成し、36.6% mAPを維持。</p>
<div class="card-tags"><span class="tag">Hardware</span><span class="tag">Edge AI</span><span class="tag">Transformer</span></div>
<div class="card-links"><a href="https://openaccess.thecvf.com/content/ICCV2025/papers/Zhang_AHCPTQ_Accurate_and_Hardware-Compatible_Post-Training_Quantization_for_Segment_Anything_Model_ICCV_2025_paper.pdf" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://github.com/Keio-CSG/AHCPTQ" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a></div>
</div>
</article>
<article class="project-card" data-tags="CIM, Hardware" data-venue="ICCAD" data-highlight="true">
<div class="card-teaser">
<img src="assets/images/pacim-teaser.svg" alt="PACiM: A Sparsity-Centric Hybrid Compute-in-Memory Architecture via Probabilistic Approximation teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-iccad">ICCAD 2024</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">PACiM</h2>
<p class="card-authors"><strong>W. Zhang</strong>, <strong>S. Ando</strong>, <strong>Y.C. Chen</strong>, <strong>S. Miyagi</strong>, S. Takamaeda-Yamazaki, <strong>K. Yoshioka</strong></p>
<p class="card-description">A sparsity-centric hybrid compute-in-memory architecture using probabilistic approximation (PAC)
to convert vector operations into scalar computations, halving memory transfers by eliminating LSB transmission.
Achieves 14.63 TOPS/W in 65 nm CMOS with 4× lower approximation error than prior methods,
maintaining ResNet-18 ImageNet accuracy.</p>
<p class="card-description-ja">確率的近似（PAC）によりベクトル演算をスカラー演算に変換し、スパース性を最大限活用するハイブリッドCIMアーキテクチャ。最小有効ビットの省略でメモリ転送量を半減し、従来比4倍の近似誤差低減を実現。65nm CMOSで14.63 TOPS/Wを達成し、ResNet-18でImageNet精度を維持する。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span></div>
<div class="card-links"><a href="https://dl.acm.org/doi/10.1145/3676536.3676704" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://github.com/Keio-CSG/PACiM" class="link-badge link-code" target="_blank" rel="noopener" aria-label="Code"><span class="badge-icon"><svg role="img" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg></span><span class="badge-label">Code</span></a></div>
</div>
</article>
<article class="project-card" data-tags="CIM, Hardware" data-venue="ASP-DAC" data-highlight="false">
<div class="card-teaser">
<img src="assets/images/osa-hcim-teaser.svg" alt="OSA-HCIM: On-The-Fly Saliency-Aware Hybrid SRAM CIM with Dynamic Precision Configuration teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-aspdac">ASP-DAC 2024</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">OSA-HCIM</h2>
<p class="card-authors"><strong>Y.C. Chen</strong>, <strong>S. Ando</strong>, D. Fujiki, S. Takamaeda-Yamazaki, <strong>K. Yoshioka</strong></p>
<p class="card-description">A hybrid SRAM compute-in-memory macro with on-the-fly saliency detection
and dynamic precision configuration.
Reduces computation by skipping non-salient activations while maintaining
inference accuracy through adaptive bit-width selection.</p>
<p class="card-description-ja">オンザフライ顕著性検出と動的精度設定を備えたハイブリッドSRAMコンピュートインメモリマクロ。非顕著な活性化をスキップしつつ適応的なビット幅選択で精度を維持する。エッジデバイスでの高効率・高精度推論の両立を実現する新しい設計手法を提案する。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span></div>
<div class="card-links"><a href="https://arxiv.org/abs/2308.15040" class="link-badge link-arxiv" target="_blank" rel="noopener" aria-label="arXiv"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><text x="1" y="17" font-family="serif" font-size="13" font-weight="bold" fill="white">ar</text></svg></span><span class="badge-label">arXiv</span></a></div>
</div>
</article>
<article class="project-card" data-tags="CIM, Hardware, Edge AI" data-venue="ISSCC" data-highlight="true">
<div class="card-teaser">
<img src="assets/images/cap-reconfigured-cim-teaser.svg" alt="A 818-4094 TOPS/W Capacitor-Reconfigured CIM Macro for Unified Acceleration of CNNs and Transformers teaser" loading="lazy" />
</div>
<div class="card-body">
<div class="card-meta">
<span class="badge badge-venue badge-isscc">ISSCC 2024</span>
<span class="badge badge-status badge-published">Published</span>
</div>
<h2 class="card-title">CR-CIM</h2>
<p class="card-authors"><strong>K. Yoshioka</strong></p>
<p class="card-description">A capacitor-reconfigured compute-in-memory macro achieving 818–4094 TOPS/W
for unified acceleration of both CNNs and Transformers.
Dynamic capacitor reconfiguration adapts analog compute precision to each layer's bit-width requirements,
enabling the first analog CIM to handle Transformer inference while delivering 10× energy efficiency
over prior work in CNN mode at 4094 TOPS/W.</p>
<p class="card-description-ja">CNNとTransformerの両モデルを統一的に加速する818〜4094 TOPS/Wのキャパシタ再構成型コンピュートインメモリマクロ。層ごとのビット幅要件に合わせてキャパシタ構成を動的に切り替えることで、幅広いモデルアーキテクチャに対応したオンチップ推論を実現する。CNNモードでは従来比10倍のエネルギー効率4094 TOPS/Wを達成。</p>
<div class="card-tags"><span class="tag">CIM</span><span class="tag">Hardware</span><span class="tag">Edge AI</span></div>
<div class="card-links"><a href="https://ieeexplore.ieee.org/document/10689660" class="link-badge link-paper" target="_blank" rel="noopener" aria-label="Paper"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8l-6-6zm-1 1.5L18.5 9H13V3.5zM8.5 15.5h-1v-4h1.3c.8 0 1.2.5 1.2 1.1 0 .7-.5 1.1-1.3 1.1H8.5v1.8zm0-2.6h.2c.3 0 .5-.1.5-.4 0-.3-.2-.4-.5-.4H8.5v.8zm3.5 2.6h-1v-4h1c1.1 0 1.8.7 1.8 2s-.7 2-1.8 2zm0-3.2h-.1v2.4h.1c.5 0 .9-.3.9-1.2 0-.9-.4-1.2-.9-1.2zm3.5 3.2h-1v-4h2.3v.8H15v.9h1.2v.7H15v1.6z"/></svg></span><span class="badge-label">Paper</span></a><a href="https://www.jst.go.jp/pr/announce/20240219/" class="link-badge link-press" target="_blank" rel="noopener" aria-label="Press"><span class="badge-icon"><svg viewBox="0 0 24 24" fill="currentColor" xmlns="http://www.w3.org/2000/svg"><path d="M4 22a2 2 0 0 1-2-2V4a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2v3h2v13a2 2 0 0 1-2 2H4zm14-2h2V9h-2v11zM4 4v16h12V4H4zm2 4h8v2H6V8zm0 4h8v2H6v-2zm0 4h5v2H6v-2z"/></svg></span><span class="badge-label">Press</span></a></div>
</div>
</article>
</div>
</section>
</main>
</div>
<footer>
<div class="container">
<span>© 2026 Keio Computing and Sensing Group</span>
<a href="https://sites.google.com/keio.jp/keio-csg" target="_blank" rel="noopener">sites.google.com/keio.jp/keio-csg</a>
</div>
</footer>
<script>
(function () {
const tagBar = document.getElementById('filter-tag');
const venueBar = document.getElementById('filter-venue');
const hlToggle = document.getElementById('toggle-highlight');
const grid = document.getElementById('project-grid');
let activeTag = 'all';
let activeVenue = 'all';
let activeHighlight = false;
function applyFilters() {
grid.querySelectorAll('.project-card').forEach(function (card) {
const tags = (card.dataset.tags || '').split(',').map(function (t) { return t.trim(); });
const venue = (card.dataset.venue || '').trim();
const highlight = card.dataset.highlight === 'true';
const tagOk = activeTag === 'all' || tags.includes(activeTag);
const venueOk = activeVenue === 'all' || venue === activeVenue;
const highlightOk = !activeHighlight || highlight;
card.classList.toggle('hidden', !(tagOk && venueOk && highlightOk));
});
grid.querySelectorAll('.project-section').forEach(function (sec) {
sec.style.display = sec.querySelectorAll('.project-card:not(.hidden)').length > 0 ? '' : 'none';
});
}
tagBar.addEventListener('click', function (e) {
const btn = e.target.closest('.filter-btn');
if (!btn) return;
tagBar.querySelectorAll('.filter-btn').forEach(function (b) { b.classList.remove('active'); });
btn.classList.add('active');
activeTag = btn.dataset.tag;
applyFilters();
});
venueBar.addEventListener('click', function (e) {
const btn = e.target.closest('.filter-btn');
if (!btn) return;
venueBar.querySelectorAll('.filter-btn').forEach(function (b) { b.classList.remove('active'); });
btn.classList.add('active');
activeVenue = btn.dataset.venue;
applyFilters();
});
hlToggle.addEventListener('click', function () {
activeHighlight = !activeHighlight;
hlToggle.classList.toggle('active', activeHighlight);
applyFilters();
});
})();
</script>
</body>
</html>