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
Mip-NeRF 360: Unbounded Anti-Aliased Neural Rad...
Search
peisuke
April 30, 2023
Technology
0
9.4k
Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields
第58回 コンピュータビジョン勉強会@関東
深層学習+3D論文読み会
Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields
peisuke
April 30, 2023
Tweet
Share
More Decks by peisuke
See All by peisuke
LangGraphで始めるマルチエージェントシステム
peisuke
13
4k
AutoGenで作るLLM Agen
peisuke
1
1.4k
Self-RAG: Learning to Retrieve, Generate and Critique through Self-Reflections
peisuke
9
1.4k
LangChain Toolsの運用と改善
peisuke
4
2.6k
GNeRF: GAN-based Neural Radiance Field without Posed Camera
peisuke
1
730
TTS Skins: Speaker Conversion via ASR
peisuke
0
370
A Quantum Computational Approach to Correspondence Problems on Point Sets
peisuke
0
670
F0-Consistent Many-to-many Non-parallel Voice Conversion via Conditional Autoencoder
peisuke
0
180
YOLACT real-time instance segmentation
peisuke
1
260
Other Decks in Technology
See All in Technology
オプトインカメラ:UWB測位を応用したオプトイン型のカメラ計測
matthewlujp
0
170
AWS re:Invent 2024 ふりかえり
kongmingstrap
0
130
KubeCon NA 2024 Recap: How to Move from Ingress to Gateway API with Minimal Hassle
ysakotch
0
200
宇宙ベンチャーにおける最近の情シス取り組みについて
axelmizu
0
110
GitHub Copilot のテクニック集/GitHub Copilot Techniques
rayuron
26
11k
Amazon VPC Lattice 最新アップデート紹介 - PrivateLink も似たようなアップデートあったけど違いとは
bigmuramura
0
190
re:Invent をおうちで楽しんでみた ~CloudWatch のオブザーバビリティ機能がスゴい!/ Enjoyed AWS re:Invent from Home and CloudWatch Observability Feature is Amazing!
yuj1osm
0
120
生成AIのガバナンスの全体像と現実解
fnifni
1
180
株式会社ログラス − エンジニア向け会社説明資料 / Loglass Comapany Deck for Engineer
loglass2019
3
31k
マイクロサービスにおける容易なトランザクション管理に向けて
scalar
0
110
Amazon SageMaker Unified Studio(Preview)、Lakehouse と Amazon S3 Tables
ishikawa_satoru
0
150
10個のフィルタをAXI4-Streamでつなげてみた
marsee101
0
160
Featured
See All Featured
The Straight Up "How To Draw Better" Workshop
denniskardys
232
140k
Rebuilding a faster, lazier Slack
samanthasiow
79
8.7k
実際に使うSQLの書き方 徹底解説 / pgcon21j-tutorial
soudai
169
50k
The Art of Programming - Codeland 2020
erikaheidi
53
13k
Building Adaptive Systems
keathley
38
2.3k
Six Lessons from altMBA
skipperchong
27
3.5k
Cheating the UX When There Is Nothing More to Optimize - PixelPioneers
stephaniewalter
280
13k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
29
2k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
159
15k
Navigating Team Friction
lara
183
15k
How GitHub (no longer) Works
holman
311
140k
The Success of Rails: Ensuring Growth for the Next 100 Years
eileencodes
44
6.9k
Transcript
Mip-NeRF 360: Unbounded Anti- Aliased Neural Radiance Fields Authors: Jonathan
T. Barron, Ben Mildenhall, Dor Verbin Presenter: @peisuke
ࣗݾհ ໊લɿ౻ຊ ܟհ ॴଐɿ"#&+"ʢ-BCTϥϘʣ ෭ۀɿ.-Λத৺ʹ૬ஊ͔Β։ൃ·Ͱͬͯ·͢ ੋඇ͓͕͚Լ͍͞ ׆ಈɿ5XJUUFS!QFJTVLF (JUIVC IUUQTHJUIVCDPNQFJTVLF 2JJUB
IUUQTRJJUBDPNQFJTVLF 4MJEF4IBSF IUUQTXXXTMJEFTIBSFOFU'VKJNPUP,FJTVLF 4QFBLFS%FDL IUUQTTQFBLFSEFDLDPNQFJTVLF গ͚ͩ͠ॻ͍ͯ·͢
֓ཁ • ͲΜͳจʁ • /F3'ͰϘϦϡʔϜϨϯμϦϯάʹ͓͍֤ͯͷ৭ɾີΛ//Ͱ ग़ྗ͢Δ͜ͱͰը૾Λੜ • /F3'ʹ͓͍ͯɺத৺͔Βԕ͍େͳۭؒΛ//Ͱ࠶ݱ͢Δ͜ͱ͕͠ ͘ɺத৺Ҏ֎ͷ෮ݩ͕͔ͬͨ͠Λղܾ •
ʁ • ۭؒΛมͯ͠ԕ͍ྖҬΛখ͍͞ྖҬʹຒΊࠐΉ͜ͱͰԕ͞ͷΛ ղܾ • ͦͷଞʹɺৠཹʹΑΔߴԽɾਖ਼نԽʹΑΔϊΠζݮͳͲΛ͍ͯ͠ Δ
ख๏ͷσϞ
/F3'ͷ֓ཁʢʣ • ۭؒͷʢҐஔ͖ͷ࣍ݩʣʹؔ͢Δ৭ɾີΛ༧ଌ͢Δ ωοτϫʔΫ • ΧϝϥͷҐஔΛܾΊΔͱɺωοτϫʔΫͷग़ྗ͢ΔΛݩʹϘ ϦϡʔϜϨϯμϦϯά͢Δ͜ͱͰը૾Λੜ 1PTJUJPOBM&ODPEJOH͞Εͨͷ ใ ࣍ϖʔδͰઆ໌
ີ͔Β XΛܭࢉ
/F3'ͷ֓ཁʢʣ • ۭؒͷΛߴ࣍ݩͷಛྔʹมʢQPTJUJPOBMFODPEJOHɺ 1&ʣ͢Δ͜ͱͰߴਫ਼Խ 𝛾 𝒙 = sin 2!𝒙𝜋 ,
cos 2!𝒙𝜋 , … , sin 2"𝒙𝜋 , cos 2"𝒙𝜋
NJQ/F3'ͷ֓ཁʢʣ • 𝛾 𝒙 Λͱͯ͠ѻ͍ɺͦͷҐஔͷ৭ΛϨϯμϦϯά͢Δͱɺ पғͷใΛಘΒΕͣδϟΪʔ͕ൃੜ͢Δ • ࢹઢΛίʔϯͰଊ͑ɺίʔϯͷੵΛ༻͍ΔࣄͰߴਫ਼Խ • ίʔϯΛΨεͰۙࣅɺ1&ͷظΛ͏͜ͱͰɺੵΛ
؆қతʹܭࢉ Ͱͳ͘ɺ֤ྖҬͷ 1&ͷظΛ͏
NJQ/F3'ͷ֓ཁʢʣ • 1&ͨ͠ಛྔͷظΛऔΔͱɺࢹ͔Βԕ͍ྖҬͰߴपྖ Ҭ͕খ͘͞ͳΔ • ڑʹԠͨ͡ૈ͞ΛՃຯͯ͠ωοτϫʔΫ͕৭Λग़ྗͯ͘͠ΕΔ ࢹۙ͘Ͱɺ͕ ڱ͍ͨΊߴपྖҬͷ ใ͕Δ ࢹ͔ΒΕΔͱɺ
͕͘ͳΔͨΊप ͷΈʹͳΔ
NJQ/F3'ͷ՝ • Χϝϥ͔Βԕ͍ྖҬίʔϯ ͕େ͖͘ͳΔ • த৺ʹ͋ΔମҎ֎ͷपғ ڥͷ෮ݩ͕ࠔ த৺ྖҬ΄Ͳࡉ͔ ͍ใ͕ඞཁ पғͷใૈͯ͘ྑ͍
֎ଆͷྖҬͷσʔλ͕ૄʹͳΔͨΊɺ ্ख͘෮ݩͰ͖ͳ͍
4DFOFBOE3BZ1BSBNFUFSJ[BUJPO • த৺͔ΒܘͷྖҬ·ͰมΘΒͣ • ܘҎ্ͷྖҬɺԕ͘ͳΔ΄ͲڑΛॖΊɺܘͷྖҬʹ ແݶԕํ͕ऩ·ΔΑ͏ʹม • ࣍ۙࣅͯ͠ϠίϏߦྻΛ͍ΨεΛม ਤͷପԁҎԼͰม
ৠཹΛ༻͍ͨߴԽʢʣ • NJQ/F3'Ͱɺ·ͣશҬΛૈཻ͍ͰϨΠτϨʔγϯά͠ɺମ͕Կ ॲʹ͋Δͷ͔Λେ·͔ʹѲ • ࣍ʹɺମ͕ଘࡏ͢ΔҐஔʹ੍ݶ͠ɺ࠶ϨΠτϨʔγϯά • ͜ΕΛ܁Γฦ͢͜ͱͰৄࡉͳܗঢ়Λ෮ݩՄೳɺܭࢉ͕͍ͱ͍͏՝͕͋ Δ
ৠཹΛ༻͍ͨߴԽʢ̎ʣ • ૈ͍ྖҬ୳ࡧϑΣʔζͰɺ৭ͷใΛ࣋ͨͳ͍ɺΑΓখ͞ͳϞσϧ ʢ1SPQ.-1ʣͰମͷҐஔΛ୳ࡧ • ࠷ޙͷϨϯμϦϯάͷΈɺ৭ͷใΛ࣋ͭϞσϧʢ/F3'.-1ʣͰ୳ࡧ • ֤ϞσϧʹҰ؏ੑΛ࣋ͨͤΔͨΊɺωοτϫʔΫͷৠཹΛ༻͍Δ
ωοτϫʔΫͷৠཹʹ͍ͭͯ • /F3'.-1͔ΒಘΒΕΔମͷҐஔͷใΛ༻͍ͯɺ1SPQ.-1Λֶश ͢Δ͜ͱͰҰ؏ੑΛอͭ • ྆ํͷωοτϫʔΫؒͰඞཁͱͳΔCJOͷີҟͳΔͷͰ Λൺֱͨ͠ৠཹ͕ग़དྷͳ͍ • /F3'.-1͔ΒಘΒΕΔॏΈͷώετάϥϜΑΓɺ1SPQ.-1ͷώε τάϥϜ͕େ͖͘ͱ͍͏੍Λ͏
• ͜ΕͪΐͬͱෳࡶͳͷͰ࣍ͷϖʔδͰઆ໌ ࠷ऴతʹɺମྖҬʹ͓ ͍ͯX͕େ͖͘ͳΔɻ· ͨɺૈ͍CJOͷ1SPQ.-1 ͕ಘΒΕΔ
• ϨΠ্ʹ͋ΔମͷॏΈXͷώετάϥϜΛߟ͑Δ ώετάϥϜͷൺֱ ϨΠ্ͷମͷॏΈ w ମͷॏΈͷώετάϥϜ
ώετάϥϜͷൺֱ /F3'.-1ͷ ώετάϥϜ 1SPQ.-1ͷ ώετάϥϜ /F3'.-1ͷͱ͋ΔྖҬ5JͷॏΈXJ ΑΓɺ5JΛؚΉྖҬ5Kͷώετά ϥϜXKͷ໘ੵͷํ͕େ͖͘ͳΔ
ਖ਼ଇԽ • /F3'Λֶश͢Δͱɺಛʹԕ͍ΤϦΞʹ͓͍ͯܗঢ়෮ݩͷͨ Ίͷใ͕Γͣෆྑઃఆ͕ൃੜ͠ɺϊΠζ͕ൃੜ • ਖ਼ଇԽʹΑͬͯϊΠζͷݮΛߦ͏
ਖ਼ଇԽ • ॏΈ͆ͷ͕ͳΔ͘σϧλؔʹۙ͘ͳΔΑ͏ͳ੍Λͭ ͚Δ σϧλؔʹͳΔͱɺ X V ɺX W ͕ڞʹͷ࣌
ʹɺVW͕ʹ͖ۙͮɺͦ ΕҎ֎ɺΑ্ͬͯࣜ θϩʹۙͮ͘ 𝑢 ≠ 𝑣ͷ࣌ʹɺw 𝑢 , 𝑤 𝑣 > 0 Ͱ͋ΔͱϖφϧςΟͱͳΔ
ਖ਼ଇԽ • ઌఔͷੵܭࢉͰ͖ͳ͍͕ɺX֤ΠϯλʔόϧͰҰఆ ʹͳΔͷͰɺҎԼͷΑ͏ʹॻ͖දͤΔ • ޯΛܭࢉ͢ΔͱɺਤͷΑ͏ͳํʹॏΈ͕มԽ͢Δྗ͕ಇ͘
࠷దԽ • ը૾ͷ෮ݩϩεɺਖ਼ଇԽʹΑͬͯ/F3'.-1Λɺ࠷ޙͷ߲Ͱ 1SPQ.-1Λֶश • Ϟσϧ • 1SPQ.-1ύϥϝʔλͷӅΕΛ࣋ͭɺϨΠϠʔͷ.-1 • /F3'.-1ύϥϝʔλͷӅΕΛ࣋ͭɺϨΠϠʔͷ.-1
ը૾ͷ෮ݩϩε ਖ਼ଇԽ߲ 1SPQ.-1ͷֶश
࣮ݧ݁Ռ • σʔλ • ৽͍͠σʔληοτɺͭͷ֎ɺͭͷ • ݅ • /F3' NJQ/F3'
/F3' ͱൺֱ • $0-."1ͰࣗݾҐஔਪఆ • ܭࢉڥ • 516WXJUIDPSFT • ࢦඪ • 14/3 44*. -1*14
ఆྔධՁ
ఆੑධՁ
·ͱΊ • NJQ/F3'ͷ՝Ͱ͋ͬͨԕํͷྖҬΛਫ਼ྑ͘෮ݩ͢Δख๏ ΛఏҊ • 4DFOFBOE3BZ1BSBNFUFSJ[BUJPOʹΑΔԕํ෮ݩߴਫ਼Խɺ ৠཹʹΑΔߴԽɺਖ਼ଇԽʹΑΔϊΠζআڈΛఏҊ