Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
三次元形状とディープラーニング
Search
Tatsuya Yatagawa
December 21, 2020
Technology
310
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
三次元形状とディープラーニング
三次元形状処理とディープラーニングの初歩についてまとめたスライドです。
Tatsuya Yatagawa
December 21, 2020
More Decks by Tatsuya Yatagawa
See All by Tatsuya Yatagawa
Event-based Camera Simulation using Monte Carlo Path Tracing with Adaptive Denoising (ICIP 2023)
tatsy
0
570
VC'23 チュートリアル講演 『三次元メッシュと深層学習』
tatsy
0
240
SIGGRAPH 2022: Neural Geometry Processing
tatsy
0
260
Real-time Shading with Free-form Planar Area Lights using Linearly Transformed Cosines (JCGT & I3D 2022)
tatsy
0
1k
SIGGRAPH Asia 2021: Real-Time Rendering
tatsy
0
150
SIGGRAPH 2021: Denoising and Guiding
tatsy
0
120
SIGGRAPH 2021: Shape Generation
tatsy
0
200
SIGGRAPH 2021: Meshing
tatsy
0
130
12. Keyboard & Mouse
tatsy
0
240
Other Decks in Technology
See All in Technology
知見・人・API・DB・予算 ─ ナイナイ尽くしだった人事データ整備 with dbt、5年間の学び
ken6377
1
170
DMM.com 購入改善推進チーム におけるCodeRabbitを用いた レビューフロー改善の一例
ysknsid25
1
520
SRE歴2ヶ月でも開発6年の知見を活かして、チームで止まっていた環境改善を前に進めた話
a_ono
1
220
ゼロをイチにする仕事が終わったあと
smasato
0
310
なぜ私たちのSREプラクティスはなかなか機能しないのか 〜システムより先に組織を見る〜 / Why our SRE practices aren't really working
vtryo
1
2.4k
Control Planeで育てるBtoB SaaSの認証基盤 - SRE NEXT 2026
pokohide
1
1.5k
グローバルチームと挑むプロダクト開発
sansantech
PRO
1
160
NDIAS CTF 2026 問題解説会資料
bata_24
0
180
Keeping applications secure by evolving OAuth 2.0 and OpenID Connect
ahus1
PRO
1
150
Why is RC4 still being used?
tamaiyutaro
0
310
ループエンジニアリングでE2Eテストを実践
noriyukitakei
0
310
最適な自走を最小限の支援で — M&Aで拡大する組織で少人数SREが挑んだ1年 / SRE NEXT 2026
genda
0
190
Featured
See All Featured
Lightning talk: Run Django tests with GitHub Actions
sabderemane
0
220
AI: The stuff that nobody shows you
jnunemaker
PRO
8
770
YesSQL, Process and Tooling at Scale
rocio
174
15k
Agile Actions for Facilitating Distributed Teams - ADO2019
mkilby
0
220
Design of three-dimensional binary manipulators for pick-and-place task avoiding obstacles (IECON2024)
konakalab
0
480
Ethics towards AI in product and experience design
skipperchong
2
320
A designer walks into a library…
pauljervisheath
211
24k
Lessons Learnt from Crawling 1000+ Websites
charlesmeaden
PRO
1
1.3k
Prompt Engineering for Job Search
mfonobong
0
370
Paper Plane
katiecoart
PRO
1
52k
Conquering PDFs: document understanding beyond plain text
inesmontani
PRO
4
2.9k
Designing for humans not robots
tammielis
254
26k
Transcript
None
• 機械学習 ニューラルネット 深層学習
• • • θ • = ; = 2 2
+ 1 + 0 ෩ = arg min =0 −1 ( , ; )
• • • 0 = 0 + 0 1 =
1 + 1 2 = 2 + 2 = 0 0 + 1 1 + 2 2 + = 0 0 + 1 1 + 2 2 +(0 0 + 1 1 + 2 2 + )
• • • 0 1 2 0 = ReLU(0 +
0 ) 1 = ReLU(1 + 1 ) 2 = ReLU(2 + 2 )
= 32 = 128 = 512 • •
• • = 512 = 1 = = 4
•
•
•
•
•
•
• • • • • • • • • •
•
(c) Velodyne (c) Andrew Tallon (c) Microsoft (c) Sony
• , , = 0
′ = − ′ = + ′
• ☺ ☺ ☺ ☺ ,
, = 0
☺ ☺ ☺ ☺ , ,
= 0
• • • • •
• • • •
•
•
• ⚫ ⚫ ⚫
• • •
• • •
• •
• • •
• • • • • • •
☺ ☺ ☺ ☺ , ,
= 0
• • • (c) Velodyne
• • • (0 , 0 , 0 ) (1
, 1 , 1 ) ( , , ) (−1 , −1 , −1 ) (+1 , +1 , +1 ) (0 , 0 , 0 ) (1 , 1 , 1 ) (−1 , −1 , −1 ) (+1 , +1 , +1 ) ( , , )
• • •
None
⚫
⚫
• •
• • •
•
•
•
• • • • • • •
☺ ☺ ☺ ☺ , ,
= 0
• • •
• • • •
• • ∗ = ℱ−1(ℱ ⋅ ℱ ) ′ =
ℱ−1(Θ ∘ ℱ )
• = 0 1 1 0 1 0 1 0
1 1 0 1 0 0 1 0 = − − 1 2− 1 2 = :
• • • = T
• () ′ = ℱ−1(Θ ∘ ℱ ) ′ =
ℱ = () ℱ−1 = ()
• • • +1 +2 +1 = ReLU ⋅
• • •
• • •
• •
• • •
• • •
• •
• • • • • •
☺ ☺ ☺ ☺ , ,
= 0
• , , =
• • •
• • • •
• •
• • •
• • •
• • • •
•
• • • •
• • •
• • • •
• • •
• • • •
• • • • • • • •
• • • • • • • • •