takminの書きっぱなし備忘録 @はてなブログ

発表者	論文タイトル	発表資料
alfredplpl	Back to Basics: Let Denoising Generative Models Denoise	https://www.docswell.com/s/alfredplpl/Z8N989-2026-07-18-082234
shunk031	PSDesigner: Automated Graphic Design with a Human-Like Creative Workflow	https://speakerdeck.com/shunk031/kantocv-67th-cvpr-2026
lychee1223_Lab	SciGA: A Comprehensive Dataset for Designing Graphical Abstracts in Academic Papers Paper2Figure: A Multi-Agent Collaborative System for Figure Generation Towards Academic Research Paper	https://speakerdeck.com/lychee1223/kantocv-67th-cvpr-2026
kzykmyzw	SoccerMaster: A Vision Foundation Model for Soccer Understanding	https://speakerdeck.com/kzykmyzw/soccermaster-a-vision-foundation-model-for-soccer-understanding
Ito	What Is the Optimal Ranking Score Between Precision and Recall? We Can Always Find It and It Is Rarely F1	https://speakerdeck.com/keiichiito1978/20260618-cvprdu-mihui
welldone	Sensor2Sensor: Cross-Embodiment Sensor Conversion for Autonomous Driving	https://speakerdeck.com/welldone/cvmian-qiang-hui-sensor2sensor
nasnetou	Can You Learn to See Without Images? Procedural Warm-Up for Vision Transformers	https://docs.google.com/presentation/d/13WFHhwXBmyhoa1bDJyeU8CvrrTLb-yZAacYtwr3eiw4/
Kenji	AT-VLA: Adaptive Tactile Injection for Enhanced Feedback Reaction in Vision-Language-Action Models	https://speakerdeck.com/tsukamotokenji/at-vla-b09cef8c-8208-4d1c-82e2-6cf35ec1ac84
TatsuyaSuzuki	RoboWheel: A Data Engine from Real-World Human Demonstrations for Cross-Embodiment Robotic Learning	https://speakerdeck.com/x_ttyszk/di-67hui-konpiyutabiziyonmian-qiang-hui-lun-wen-shao-jie-robowheel-a-data-engine-from-real-world-human-demonstrations-for-cross-embodiment-robotic-learning

発表者

論文タイトル

発表資料

alfredplpl

Back to Basics: Let Denoising Generative Models Denoise

https://www.docswell.com/s/alfredplpl/Z8N989-2026-07-18-082234

shunk031

PSDesigner: Automated Graphic Design with a Human-Like Creative Workflow

https://speakerdeck.com/shunk031/kantocv-67th-cvpr-2026

lychee1223_Lab

SciGA: A Comprehensive Dataset for Designing Graphical Abstracts in Academic Papers
Paper2Figure: A Multi-Agent Collaborative System for Figure Generation Towards Academic Research Paper

https://speakerdeck.com/lychee1223/kantocv-67th-cvpr-2026

kzykmyzw

SoccerMaster: A Vision Foundation Model for Soccer Understanding

https://speakerdeck.com/kzykmyzw/soccermaster-a-vision-foundation-model-for-soccer-understanding

Ito

What Is the Optimal Ranking Score Between Precision and Recall? We Can Always Find It and It Is Rarely F1

https://speakerdeck.com/keiichiito1978/20260618-cvprdu-mihui

welldone

Sensor2Sensor: Cross-Embodiment Sensor Conversion for Autonomous Driving

https://speakerdeck.com/welldone/cvmian-qiang-hui-sensor2sensor

nasnetou

Can You Learn to See Without Images? Procedural Warm-Up for Vision Transformers

https://docs.google.com/presentation/d/13WFHhwXBmyhoa1bDJyeU8CvrrTLb-yZAacYtwr3eiw4/

Kenji

AT-VLA: Adaptive Tactile Injection for Enhanced Feedback Reaction in Vision-Language-Action Models

https://speakerdeck.com/tsukamotokenji/at-vla-b09cef8c-8208-4d1c-82e2-6cf35ec1ac84

TatsuyaSuzuki

RoboWheel: A Data Engine from Real-World Human Demonstrations for Cross-Embodiment Robotic Learning

https://speakerdeck.com/x_ttyszk/di-67hui-konpiyutabiziyonmian-qiang-hui-lun-wen-shao-jie-robowheel-a-data-engine-from-real-world-human-demonstrations-for-cross-embodiment-robotic-learning

2月8日は、第66回コンピュータビジョン勉強会＠関東「世界モデル論文読み会」を、チューリング株式会社様に会場をお借りして行いました。

以下、自分で見返すために資料やリンク等をまとめておきます。

登録サイト

kantocv.connpass.com

posfie

posfie.com

Youtube

今回はネットワークが不安定だったため、動画が途切れ途切れになってしまい、申し訳ございません。

www.youtube.com

コンピュータビジョン勉強会＠関東

sites.google.com

資料まとめ

発表者	論文タイトル	発表資料
tomoaki_teshima	World Models (2018) / Dreamer (2019)	Dropbox - kantocv_world_models.pdf - Simplify your life
Keiichi-Ito	WorldPrediction: A Benchmark for High-level World Modeling and Long-horizon Procedural Planning	20260208_第66回コンピュータビジョン勉強会 - Speaker Deck
Hidehisa Arai	世界モデルにおける分布外データ対応の方法論	世界モデルにおける分布外データ対応の方法論 - Speaker Deck
Godel	自律移動ロボットはWorld Modelsの夢は見るか?
大政孝充	GWM: Towards Scalable Gaussian World Models for Robotic Manipulation	第66回コンピュータビジョン勉強会@関東における株式会社ウェブファーマー大政の発表 GWMモデル \| PPTX
caprest	世界モデルで物を掴めるようになるのか？	世界モデルで物を掴めるようになるのか？.pdf - Google ドライブ
Kento Sasaki	Epona: Autoregressive Diffusion World Model for Autonomous Driving (ICCV 2025)	第66回コンピュータビジョン勉強会＠関東 Epona: Autoregressive Diffusion World Model for Autonomous Driving - Speaker Deck
takmin	Cosmos World Foundation Model Platform for Physical AI	Cosmos World Foundation Model Platform for Physical AI - Speaker Deck
Shin-kyoto	Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail	[CV勉強会@関東世界モデル論文読み会] VLA自動運転model Alpamayo-R1 - Speaker Deck
abemii	Orbis: Overcoming Challenges of Long-Horizon Prediction in Driving World Models	[CV勉強会@関東 World Model 読み会] Orbis: Overcoming Challenges of Long-Horizon Prediction in Driving World Models (Mousakhan+, NeurIPS 2025) - Speaker Deck

takmin発表資料

speakerdeck.com

雪であまり人が集まらないかと心配しましたが、とても盛況で安心しました。

発表者	発表内容	資料
abemii	World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model	https://speakerdeck.com/abemii/cvmian-qiang-hui-at-guan-dong-cvpr2025-du-mihui-world4drive-end-to-end-autonomous-driving-via-intention-aware-physical-latent-world-model-zheng-plus-iccv-2025
caprest	Fine-Grained Evaluation of Large Vision-Language Models in Autonomous Driving	https://docs.google.com/presentation/d/10JJiQdOwoKiw--6g7B5Mc2N2Y4BMpNKDCkQ5oSx5NZA
Shin	End-to-End Driving with Online Trajectory Evaluation via BEV World Model	https://speakerdeck.com/shinkyoto/cvmian-qiang-hui-at-guan-dong-iccv2025-wote-end-to-end-driving-with-online-trajectory-evaluation-via-bev-world-model
Keiichi-Ito	AnimeGamer: Infinite Anime Life Simulation with Next Game State Prediction	https://speakerdeck.com/keiichiito1978/20250916-di-65hui-konpiyutabiziyonmian-qiang-hui
alfredplpl	ICCV 2025の動画生成を眺めてみた StreamDiffusion: A Pipeline-level Solution forReal-Time Interactive Generation VACE: All-in-OneVideo Creation and Editing FiVE: A Fine-grainedVideo EditingBenchmark for EvaluatingEmerging Diffusion and Rectified Flow Models	https://www.docswell.com/s/alfredplpl/5VM392-2025-11-08-194529
ShuN Makino	Is CLIP ideal? No. Can we fix it? Yes!	https://speakerdeck.com/shun6211/lun-wen-shao-jie-is-clip-ideal-no-can-we-fix-it-yes-di-65hui-konpiyutabiziyonmian-qiang-hui-at-guan-dong
Kenji	CoA-VLA: Improving Vision-Language-Action Models via Visual-Textual Chain-of-Affordance	https://speakerdeck.com/tsukamotokenji/di-65hui-konpiyutabiziyonmian-qiang-hui
peisuke	Moto: Latent Motion Token as the Bridging Language for Learning Robot Manipulation from Videos	https://speakerdeck.com/peisuke/moto-latent-motion-token-as-the-bridging-language-for-learning-robot-manipulation-from-videos

発表者

発表内容

資料

abemii

World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model

https://speakerdeck.com/abemii/cvmian-qiang-hui-at-guan-dong-cvpr2025-du-mihui-world4drive-end-to-end-autonomous-driving-via-intention-aware-physical-latent-world-model-zheng-plus-iccv-2025

caprest

Fine-Grained Evaluation of Large Vision-Language Models in Autonomous Driving

https://docs.google.com/presentation/d/10JJiQdOwoKiw--6g7B5Mc2N2Y4BMpNKDCkQ5oSx5NZA

Shin

End-to-End Driving with Online Trajectory Evaluation via BEV World Model

https://speakerdeck.com/shinkyoto/cvmian-qiang-hui-at-guan-dong-iccv2025-wote-end-to-end-driving-with-online-trajectory-evaluation-via-bev-world-model

Keiichi-Ito

AnimeGamer: Infinite Anime Life Simulation with Next Game State Prediction

https://speakerdeck.com/keiichiito1978/20250916-di-65hui-konpiyutabiziyonmian-qiang-hui

alfredplpl

ICCV 2025の動画生成を眺めてみた
StreamDiffusion: A Pipeline-level Solution forReal-Time Interactive Generation
VACE: All-in-OneVideo Creation and Editing
FiVE: A Fine-grainedVideo EditingBenchmark for EvaluatingEmerging Diffusion and Rectified Flow Models

https://www.docswell.com/s/alfredplpl/5VM392-2025-11-08-194529

ShuN Makino

Is CLIP ideal? No. Can we fix it? Yes!

https://speakerdeck.com/shun6211/lun-wen-shao-jie-is-clip-ideal-no-can-we-fix-it-yes-di-65hui-konpiyutabiziyonmian-qiang-hui-at-guan-dong

Kenji

CoA-VLA: Improving Vision-Language-Action Models via Visual-Textual Chain-of-Affordance

https://speakerdeck.com/tsukamotokenji/di-65hui-konpiyutabiziyonmian-qiang-hui

peisuke

Moto: Latent Motion Token as the Bridging Language for Learning Robot Manipulation from Videos

https://speakerdeck.com/peisuke/moto-latent-motion-token-as-the-bridging-language-for-learning-robot-manipulation-from-videos

発表者	発表内容	資料
takmin	R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization	https://speakerdeck.com/takmin/r-score-revisiting-scene-coordinate-regression-for-robust-large-scale-visual-localization
Takeo Shibata	MotionPro: A Precise Motion Controller for Image-to-Video Generation
abemii	MegaSaM: Accurate, Fast, and Robust Structure and Motion from Casual Dynamic Videos	https://speakerdeck.com/abemii/cvmian-qiang-hui-at-guan-dong-cvpr2025-du-mihui-megasam-accurate-fast-and-robust-structure-and-motion-from-casual-dynamic-videos-li-plus-cvpr2025
s_aiueo32	Linguistics-aware Masked Image Modeling for Self-supervised Scene Text Recognition	https://speakerdeck.com/s_aiueo32/cvpr2025lun-wen-du-mihui-linguistics-aware-masked-image-modeling-for-self-supervised-scene-text-recognition
kzykmyzw	Gaze-LLE: Gaze Target Estimation via Large-Scale Learned Encoders	https://speakerdeck.com/kzykmyzw/gaze-lle-gaze-target-estimation-via-large-scale-learned-encoders
Kenji	RoboSpatial: Teaching Spatial Understanding to 2D and 3D Vision-Language Models for Robotics	https://speakerdeck.com/tsukamotokenji/di-64hui-konpiyutabiziyonmian-qiang-hui-at-guan-dong-hou-bian
caprest	SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment	https://docs.google.com/presentation/d/1YqppSVFJNaqXKHuhx3SZvob1ZCulix2X-o8PWM7Zfao/
frkake	Removing Reflections from RAW Photos	https://speakerdeck.com/frkake/removing-reflections-from-raw-photos
YutaKikuchi	DroneSplat: 3D Gaussian Splatting for Robust 3D Reconstruction from In-the-Wild Drone Imagery	https://speakerdeck.com/yutakikuchi_sd/cvmian-qiang-hui-at-guan-dong-dronesplat-3d-gaussian-splatting-for-robust-3d-reconstruction-from-in-the-wild-drone-imagery
Keiichi-Ito	Towards Zero‑Shot Anomaly Detection and Reasoning with Multimodal Large Language Models	https://speakerdeck.com/keiichiito1978/cvprmian-qiang-hui-hou-ban

発表者

発表内容

資料

takmin

R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization

https://speakerdeck.com/takmin/r-score-revisiting-scene-coordinate-regression-for-robust-large-scale-visual-localization

Takeo Shibata

MotionPro: A Precise Motion Controller for Image-to-Video Generation

abemii

MegaSaM: Accurate, Fast, and Robust Structure and Motion from Casual Dynamic Videos

https://speakerdeck.com/abemii/cvmian-qiang-hui-at-guan-dong-cvpr2025-du-mihui-megasam-accurate-fast-and-robust-structure-and-motion-from-casual-dynamic-videos-li-plus-cvpr2025

s_aiueo32