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
SIGGRAPH2020勉強会 "Creative Fabrication"
Search
yamdeck
August 15, 2020
Research
1
76
SIGGRAPH2020勉強会 "Creative Fabrication"
yamdeck
August 15, 2020
Tweet
Share
More Decks by yamdeck
See All by yamdeck
SIGGRAPH Asia 2020 勉強会 "Computational Holography"
yamdeck
0
69
SIGGRAPH2020勉強会 "VR Hardware"
yamdeck
1
190
“HCI Research as Problem-Solving”(CHI’16) で学ぶ What is HCI Research ?
yamdeck
0
360
Other Decks in Research
See All in Research
SkySense V2: A Unified Foundation Model for Multi-modal Remote Sensing
satai
3
310
説明可能な機械学習と数理最適化
kelicht
2
800
Attaques quantiques sur Bitcoin : comment se protéger ?
rlifchitz
0
120
"主観で終わらせない"定性データ活用 ― プロダクトディスカバリーを加速させるインサイトマネジメント / Utilizing qualitative data that "doesn't end with subjectivity" - Insight management that accelerates product discovery
kaminashi
15
18k
Akamaiのキャッシュ効率を支えるAdaptSizeについての論文を読んでみた
bootjp
1
370
Pythonでジオを使い倒そう! 〜それとFOSS4G Hiroshima 2026のご紹介を少し〜
wata909
0
1.2k
Remote sensing × Multi-modal meta survey
satai
4
670
生成的情報検索時代におけるAI利用と認知バイアス
trycycle
PRO
0
170
AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data
satai
3
660
まずはここから:Overleaf共同執筆・CopilotでAIコーディング入門・Codespacesで独立環境
matsui_528
3
1k
Mamba-in-Mamba: Centralized Mamba-Cross-Scan in Tokenized Mamba Model for Hyperspectral Image Classification
satai
3
420
離散凸解析に基づく予測付き離散最適化手法 (IBIS '25)
taihei_oki
PRO
1
650
Featured
See All Featured
jQuery: Nuts, Bolts and Bling
dougneiner
65
8.3k
The World Runs on Bad Software
bkeepers
PRO
72
12k
We Analyzed 250 Million AI Search Results: Here's What I Found
joshbly
0
390
30 Presentation Tips
portentint
PRO
1
180
No one is an island. Learnings from fostering a developers community.
thoeni
21
3.6k
How People are Using Generative and Agentic AI to Supercharge Their Products, Projects, Services and Value Streams Today
helenjbeal
1
92
Money Talks: Using Revenue to Get Sh*t Done
nikkihalliwell
0
130
The Success of Rails: Ensuring Growth for the Next 100 Years
eileencodes
47
7.9k
KATA
mclloyd
PRO
33
15k
Testing 201, or: Great Expectations
jmmastey
46
7.8k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
31
3.1k
How to Grow Your eCommerce with AI & Automation
katarinadahlin
PRO
0
84
Transcript
$SFBUJWF'BCSJDBUJPO
2 1BQFS*OGPSNBUJPO $PNQVUBUJPOBM*NBHF.BSLJOHPO.FUBMTWJB-BTFS *OEVDFE)FBUJOH 4FCBTUJBO$VDFSDB 1JPUS%JEZL )BOT1FUFS4FJEFM 7BIJE#BCBFJ .BY1MBODL*OTUJUVUFGPS*OGPSNBUJDT (FSNBOZ
6JWFSTJUBEFMMB4WJ[[FSBJUBMJBOB 4XJU[FSMBOE
3 • νλϯεςϯϨεͱ͍ͬͨۚଐͷද໘ʹࢎԽൽບΛ࡞Γɼޫͷׯবʹ Αͬͯൃ৭ͤ͞ΔʮϨʔβʔΧϥʔϚʔΩϯάʯͱݺΕΔٕज़͕͋Δ • ՃػϨʔβʔڧͳͲෳͷࣗ༝Λ࣋ͪɼௐઅ؆୯Ͱͳ͍ ʢΧϥϑϧ͔ͭߴղ૾ͳྑ͍݁ՌΛಘΒΕΔΑ͏ʹ͢Δͷͳ͓ͷ͜ͱ େมʣ ˣ •
ຊจͰɼܭଌϕʔε͔ͭσʔλυϦϒϯͰ࠷దԽͨ͠$PMPS*NBHF 3FQSPEVDUJPO8PSLGMPXΛߏஙͨ͠ • ΠϯΫδΣοτϓϦϯλʔͷΧϥʔΩϟϦϒϨʔγϣϯͷΑ͏ͳϫʔ ΫϑϩʔͰͰ͖ΔΑ͏ʹ͢Δ *OUSPEVDUJPO ϨʔβʔΧϥʔϚʔΩϯά ϨʔβʔΧϥʔϚʔΩϯάͷྫ ൃ৭ͷݪཧ
4 • ৭Λ࠶ݱ͢Δػࡐʹ͓͍ͯɼ৭ҬΛܾఆ͢Δ͜ͱෆՄܽʢΧϥʔϓϦϯλʔ৭Ҭ͕͋Δʣ • Ұํɼ৭ҬͷΈΛ༏ઌ͢Δ͋·Γɼղ૾͕ࣦΘΕͯࠔΔʢFY࠼ߴ͍͕ϐον͕ૈ͗͢Δʣ • ΑͬͯɼҎԼ̑ͭͷج४Λຬͨ͢Α͏ͳ৭Ҭͷ࠷దԽʢଟత࠷దԽʣ͕ඞཁ • $ISPNBUJDJUZ๛͔ͳ৭࠼Λ࣮ݱ •
)VFTQSFBE৭ΛภΒͳ͍Α͏ʹ • 3FTPMVUJPOղ૾ • 1FSGPSNBODFTQBDFEJWFSTJUZ • %FTJHOTQBDFEJWFSTJUZ • ྫ͑ҹػͰύονΛҹͯ͠ܭଌ͠ɼҹػͷ৭Ҭʹ߹Θͤͨ-65Λ࡞͢Δ 1SPQPTFE.FUIPE (BNVU&YQMPSBUJPOʢ৭Ҭ୳ࡧʣ .$.0(" .POUF$BSMP.VMUJPCKFDUJWF(FOFUJD"MHPSJUIN ͷ̍ΠςϨʔγϣϯ
5 • ৭Ҭ͕ܾఆ͞ΕΕɼޙී௨ͷϓϦϯλʔͱಉ༷ͷॲཧͱ ͳ͍ͬͯ͘ • (BNVU.BQQJOH৭Ҭʹ߹Θͤͯը૾Λ৭ม • $PMPS4FQFSBUJPOج४৭ 3(#ϓϦϯλʔͳΒ3 (
# ʹׂ • )BMGUPOJOHϋʔϑτʔϯҹͱಉ͡Α͏ʹೱΛ࠶ݱ • 7FDUPSJ[BUJPOϨʔβʔϚʔΩϯά࣌ઢඳըͳͷͰϥ Πϯঢ়ʹ 1SPQPTFE.FUIPE *NBHF3FQSPEVDUJPO $PMPS3FQSPEVDUJPO8PSLGMPX
6 • ϨʔβʔϚʔΩϯάӈͷࣸਅͷΑ͏ͳηοτΞοϓΛ४උ࣮ͯ͠ߦ • Ճʹ͓͍ͯ̓ͭͷύϥϝʔλʔΛར༻ • 'SFRVFODZ܁Γฦ͠प • 1PXFSڧ •
1VMTFXJEUIύϧε෯ • 4QFFEՃ෦ͷҠಈ FYNNT • -JOFDPVOUDMVTUFSͷϥΠϯ • )BUDIJOHDMVTUFSͷϥΠϯؒͷڑ • 1BTTDPVOUԿճಉҰՕॴͷϥΠϯΛඳը͢Δ͔ &YQFSJNFOU "CPVU-BTFS
7 • ࠨJOQVUJNBHF • தԝ(BNVU.BQQJOH ޙͷը૾ • ӈϨʔβʔϚʔΩϯά ʹΑΔඳը݁Ռ 3FTVMU
݁Ռ
8 1BQFS*OGPSNBUJPO $PNQVUBUJPOBM1BSRVFUSZ'BCSJDBUFE4UZMF5SBOTGFSXJUI 8PPE1JYFMT +VMJBO*TFSJOHIBVTFO .JDIBFM8FJONBOO 8FJ[IFO)VBOH .BUUIJBT#)VMMJO 6OJWFSTJUZPG#POO (FSNBOZ
9 • 1BSRVFUSZ ύʔέοτϦʔ ͱݺΕΔΛ׆༻ͨ͠දݱ͕͋Δ • ը૾Λೖྗͱ͠ɼ1BSRVFUSZʹΑΔ࠶ݱ GBCSJDBUJPO Λߦ͏ϫʔΫϑ ϩʔͷ࣮ݱ͓ΑͼۙࣅΞϧΰϦζϜͷఏҊΛߦͬͨ
.PUJWBUJPO ຊจͷϞνϕʔγϣϯͱͬͨ͜ͱ 1BSRVFUSZʹΑΔϙʔτϨʔτ࡞ͷྫ 1BSRVFUSZͷྫ
10 • ࡞ϓϩηεҎԼͷ௨Γ • ࡐΛ༻ҙ͠ɼεΩϟχϯάͯ͠ςΫενϟը૾Λ༻ҙ • λʔήοτΠϝʔδͱ߹ΘͤͯɼΈ߹ΘͤʢΧοτύλʔϯͱஔʣΛܭࢉ • ࣮ࡍʹࡐΛΧοτ͠ɼΈ߹Θͤɼݻఆ͢Δͱ 1SPQPTFE.FUIPE
'BCSJDBUJPO1SPDFTT 1SPQPTFE&OEUP&OE1SPDFTT
11 • ̎छྨͷϑΟϧλʔʢڧϑΟϧλʔɾκʔϕϧϑΟϧλʔʣΛը૾ʹରͯ͠ద༻͠ɼಛྔΛநग़ • • λʔήοτը૾ͱςΫενϟ ࡐ ͕େ͖͘ҟͳΔ৭ҬϑΟϧλʔԠΛ࣋ͪ͏ΔͷͰɼώετάϥϜϚονϯάΛ ద༻
$%'CBTFEIJTUPHSBNFRVBMJ[BUJPO • ͨͩ͠ɼ࣮ࡍʹద༻ͯ͠ΈΔͱద༻͕աͰ͋ͬͨͨΊɼݩͷλʔήοτڧͱద༻ޙͷڧΛิؒͨ͠ • • *5λʔήοτը૾ɼ*4ιʔεςΫενϟ F(I) = (wintens ∙ Fintens (I), wedge ∙ Fedge (I)) T F′ intens (IT ) = (1 − whist )Fintens (IT ) + whist Fequalize (It , IS ) 1SPQPTFE.FUIPE 'FBUVSF&YUSBDUJPO MFGUݩը૾ NJEEMFMFGUڧϑΟϧλͷΈ NJEEMFSJHIUิؒ SJHIUώετάϥϜϚονϯά
12 • λʔήοτը૾Λλʔήοτύονʹׂ͠ɼλʔήοτύονͱׂΓͯΔιʔεύονͷಛྔͷ͕ࠩখ͘͞ͳΔ Α͏ʹɼιʔεύονΛܾఆ͍͖ͯ͠·͢ • • 15λʔήοτύονɼ14ιʔεύον •
ͳ͓ɼλʔήοτը૾ͷύονͷॱ൪TBMJFODZTDPSF͔த৺͔Βͷڑʹґଘͯ͠ιʔτ͢Δ Di,j (x) = np ∑ k=1 (F(PT (pk )) − F(IS,i,ϕj (x + pk ))) 2 PS = arg mini,j,x Di,j (x) 1SPQPTFE.FUIPE $VUQBUUFSOPQUJNJ[BUJPO ࠨTBMJFODZTDPSF ӈத৺͔Βͷڑ
13 • ̎ͭͷΞϧΰϦζϜ͕ఏҊ͞Ε͍ͯΔ • PQUJPOBMͱͳ͍ͬͯͨͷͰຊൃදͰলུ 1SPQPTFE.FUIPE 1BUDITIBQFPQUJNJ[BUJPO 4IBQFPQUJNJ[BUJPOͷ̍ྫ
14 3FTVMUT ݁Ռ
15 1BQFS*OGPSNBUJPO 3PCP$VU)PUXJSF$VUUJOHXJUI3PCPUDPOUSPMMFE'MFYJCMF 3PET 4JNPO%VFOTFS 3PJ1PSBOOF #FSOIBSE5IPNBT[FXTLJ 4UFMJBO$PSPT
&5);VSJDI 6OJWFSTJUZPG)BJGB 6OJWFSTJUFEF.POUSFBM
16 • σϡΞϧϩϘοτΞʔϜΛ༻͍ ͯɼϗοτϫΠϠʔΧοτʹΑ ΔGBCSJDBUJPOख๏ͷఏҊ • ࣮ݱʹओʹ̏ͭͷλεΫͷ࣮ ߦ͕ඞཁ • ϩϘοτΞʔϜͷಈ͖ʹ
͏ܗঢ়ͷมԽ • λʔήοτܗঢ়ʹͳΔΑ ͏ͳαʔϑΣεۙࣅ • ϫΠϠʔΧοτΛߦ͏্ Ͱͷύεܗ 0WFSWJFX ຊจͷશମ૾
17 • ϩϘοτΞʔϜͷ֯RͱͦΕʹΑܾͬͯ·ΔϫΠϠʔܗঢ় ࠲ඪ YΛλʔ ήοτܗঢ়͔ΒٻΊ͍ͯ͘ • 6Lλʔήοτܗঢ়ͷ·ͩΧοτ͞Ε͍ͯͳ͍෦ͷ࠲ඪͰɼ6LʹରԠ͢ ΔY࠲ඪΛܾఆ͍ͯ͘͠ 1SPQPTFE.FUIPE
ϫΠϠʔܗঢ়ͷܾΊํ ಈ࡞ڥͱؔ࿈ύϥϝʔλ ϫΠϠʔܗঢ়ͷܾΊํ
18 • ϫΠϠʔܗঢ়ɾಈ࡞ͷ࠷దԽΛ ߦ͍ͬͯΔ • ͳ͓ɼ࠷దԽΛߦ͏্Ͱ࠷ॳ ʹϥϑͳύε͕ඞཁͱͳΔ͕ɼ ͜ͷύεϢʔβʔʹॻ͔ͤͯ ྑ͍͠ɼશࣗಈԽͰߦ͏͜ ͱͰ͖Δ
1SPQPTFE.FUIPE ϫΠϠʔಈ࡞ͷ࠷దԽ
19 • ܗͷΓग़͠Λͨ͠ΓɼΧοτͨ͠ͷ͔ΒͷΈཱͯͰ͖Δ 3FTVMUT 'BCSJDBUJPO3FTVMUT ݁Ռ