Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Speaker Deck
PRO
Sign in
Sign up for free
三次元形状とディープラーニング
Tatsuya Yatagawa
December 21, 2020
Technology
0
82
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
Tweet
Share
More Decks by Tatsuya Yatagawa
See All by Tatsuya Yatagawa
Real-time Shading with Free-form Planar Area Lights using Linearly Transformed Cosines (JCGT & I3D 2022)
tatsy
0
77
12. Keyboard & Mouse
tatsy
0
81
13. Object selection
tatsy
0
49
14. Arcball control
tatsy
0
160
11. Shading models
tatsy
0
67
8. Vertex array object
tatsy
0
75
9. Shader language
tatsy
0
130
10. Model loading
tatsy
0
80
4. Window resize
tatsy
0
81
Other Decks in Technology
See All in Technology
LINEのB2Bプラットフォームにおけるトラブルシューティング2選
line_developers
PRO
4
300
UWBを使ってみた
norioikedo
0
420
FoodTechにおける商流・金流・物流の進化/Evolution of Commercial, Financial, and Logistics in FoodTech
dskst
0
410
プログラマがオブジェクト指向しても幸せになれない理由
shirayanagiryuji
0
140
組織の崩壊と再生、その中で何を考え、感じたのか。 そして本当に必要だったもの
kosako
10
4.3k
What's new in Vision
satotakeshi
0
210
MoT TechTalk #12 タクシーアプリ『GO』大規模トラフィックを捌く分析データ基盤の全容に迫る!
mot_techtalk
1
380
スタートアップと技術選定と AWS
track3jyo
PRO
2
350
Custom AppをIP制限ありのままで審査に通す方法
yusuga
0
680
JAWS-UG re:Habilitaion 報告 / JAWS-UG OITA rehabilitation
hiranofumio
0
130
機械学習システムアーキテクチャ入門 #1
asei
3
1.2k
History of the ML system in KARTE
kargo113
1
660
Featured
See All Featured
Docker and Python
trallard
27
1.6k
Statistics for Hackers
jakevdp
781
210k
How to Ace a Technical Interview
jacobian
265
21k
Bash Introduction
62gerente
597
210k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
29
4.3k
VelocityConf: Rendering Performance Case Studies
addyosmani
316
22k
How to train your dragon (web standard)
notwaldorf
58
3.9k
Support Driven Design
roundedbygravity
86
8.5k
From Idea to $5000 a Month in 5 Months
shpigford
373
44k
Embracing the Ebb and Flow
colly
73
3.4k
The Brand Is Dead. Long Live the Brand.
mthomps
46
2.7k
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
212
20k
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
• , , =
• • •
• • • •
• •
• • •
• • •
• • • •
•
• • • •
• • •
• • • •
• • •
• • • •
• • • • • • • •
• • • • • • • • •