The 3D Geometric Structure Laboratory focuses on sensing, modeling, and understanding geometric structures. Research efforts involve combining knowledge and embedding technology to improve the relationship between objects and humans. The laboratory adapts techniques from computer graphics, computer vision, machine learning, and robotics. The ultimate goal is to augment the limitations of time, space, and human capabilities to create a better world.

3D 기하 구조 연구실은 기하 구조 감지, 모델링 및 이해에 중점을 둔다. 지식 결합 및 기술 내장을 통해 사물과 인간의 관계 개선을 목표로 한다. 컴퓨터 그래픽스, 컴퓨터 비전, 머신러닝 및 로봇 공학 기술을 활용하며, 시간, 공간 및 인간 능력의 한계를 보완하여 더 나은 세상을 만드는 것을 목표로 한다.

Point2Act: Efficient 3D Distillation of Multimodal LLMs for Zero-Shot Context-Aware Grasping

3D-aware Disentangled Representation for Compositional Reinforcement Learning

Geometry-Aware Scene Configurations for Novel View Synthesis

Privacy-Preserving Visual Localization with Event Cameras

Humans as a Calibration Pattern: Dynamic 3D Scene Reconstruction from Unsynchronized and Uncalibrated Videos

Motion Synthesis with Sparse and Flexible Keyjoint Control

Event-Driven Storytelling with Multiple Lifelike Humans in a 3D scene

SceneMI: Motion In-betweening for Modeling Human-Scene Interaction

Less is More: Improving Motion Diffusion Models with Sparse Keyframes

Learning 3D Scene Analogies with Neural Contextual Scene Maps

BrepDiff: Single-stage B-rep Diffusion Model

PLT: Part-wise Latent Tokens as Adaptable Motion Priors for Physically Simulated Characters

Goal-Driven Human Motion Synthesis in Diverse Tasks

Audio-aided Character Control for Inertial Measurement Tracking

Recovering Dynamic 3D Sketches from Videos

Versatile Physics-based Character Control with Hybrid Latent Representation

Isometric Regularization for Manifolds of Functional Data

ReMP: Reusable Motion Prior for Multi-domain 3D Human Pose Estimation and Motion Inbetweening

Multi-agent Exploration with Similarity Score Map and Topological Memory

I2-SLAM: Inverting Imaging Process for Robust Photorealistic Dense SLAM

ASIMO: Agent-centric Scene representation In Multi-Object manipulation

3Doodle: Compact Abstraction of Objects with 3D Strokes

Investigating Chiral Morphogenesis of Gold Using Generative Cellular Automata

Outdoor Scene Extrapolation with Hierarchical Generative Cellular Automata

Fully Geometric Panoramic Localization

NFL: Normal Field Learning for 6-DoF Grasping of Transparent Objects

Robust Novel View Synthesis with Color Transform Module

IBL-NeRF: Image-Based Lighting Formulation of Neural Radiance Fields

MAMMOS: MApping Multiple human MOtion with Scene understanding and natural interactions

Dynamic Mesh Recovery from Partial Point Cloud Sequence

Calibrating Panoramic Depth Estimation for Practical Localization and Mapping

LDL: Line Distance Functions for Panoramic Localization

PMP: Learning to Physically Interact with Environments using Multiple Part-wise Motion Priors

Text2Scene: Text-driven Indoor Scene Stylization with Part-aware Details

Balanced Spherical Grid for Egocentric View Synthesis

Text2PointCloud: Text-Driven Stylization for Sparse PointCloud

Closed-loop optimization of nanoparticle synthesis enabled by robotics and machine learning

Ev-NeRF : Event Based Neural Radiance Field

MoDA: Map style transfer for self-supervised Domain Adaptation of embodied agents

CPO: Change Robust Panorama to Point Cloud Localization

MasKGrasp: Mask-based Grasping for Scenes with Multiple General Real-world Objects

Ev-TTA: Test-Time Adaptation for Event-Based Object Recognition

Probabilistic Implicit Scene Completion

Neural Marionette: Unsupervised Learning of Motion Skeleton and Latent Dynamics from Volumetric Video

Self-Supervised Domain Adaptation for Visual Navigation with Global Map Consistency

Progressive Growing of Points with Tree-structured Generators

Fast and Lightweight Path Guiding Algorithm on GPU

PICCOLO: Point Cloud-Centric Omnidirectional Localization

N-ImageNet: Towards Robust, Fine-Grained Object Recognition with Event Cameras

Structure-from-Sherds: Incremental 3D Reassembly of Axially Symmetric Pots from Unordered and Mixed Fragment Collections

IMAT: The Iterative Medial Axis Transform

Auto-rigging 3D Bipedal Characters in Arbitrary Poses

GATSBI: Generative Agent-centric Spatio-temporal Object Interaction

Learning to generate 3D shapes with Generative Cellular Automata

3D Pots Configuration System by Optimizing Over Geometric Constraints

SAUM: Symmetry-Aware Upsampling Module for Consistent Point Cloud Completion

Novel View Synthesis with Skip Connections

GROOT: A Real-time Streaming System of High-Fidelity Volumetric Videos

Automatic Pose Generation for Robotic 3-D Scanning of Mechanical Parts

PotSAC: A Robust Axis Estimator for Axially Symmetric Pot Fragments

RL-GAN-Net: A Reinforcement Learning Agent Controlled GAN Network for Real-Time Point Cloud Shape Completion

3D Vision Lab

교수님 소개

연구실 소개

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

Point2Act: Efficient 3D Distillation of Multimodal LLMs for Zero-Shot Context-Aware Grasping

3D-aware Disentangled Representation for Compositional Reinforcement Learning

Geometry-Aware Scene Configurations for Novel View Synthesis

연구실 보유 특허

대학별 입시끈

우리 연구실 모집

교수님 소개

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

Point2Act: Efficient 3D Distillation of Multimodal LLMs for Zero-Shot Context-Aware Grasping

3D-aware Disentangled Representation for Compositional Reinforcement Learning

Geometry-Aware Scene Configurations for Novel View Synthesis

연구실 보유 특허

최근 3년간 연구실적 높은 연구실

대학별 입시끈

우리 연구실 모집