• 목록

2022년 11/8 ~ 12/9 동안 열린 세계적인 인공지능 학술대회인 'NeurIPS 2022'에서 소개된 21편의 논문들에 대한 세미나를 진행한다.

본 세미나에서 소개하는 논문들은 Representation Learning, Motion, Imitation Learning/Reinforcement Learning, Dataset/Benchmark 4분류로 나누어 선정하였다.

각 분류별 논문 리스트는 다음과 같다.

Representation Learning

- Where to Pay Attention in Sparse Training for Feature Selection
- Cross-Modal Learning for Image-Guided Point Cloud Shape Completion
- Wavelet Feature Maps Compression for Image-to-Image CNNs
- Adapt Former: Adapting Vision Transformers for Scalable Visual Recognition

- Point Next: Revisiting PointNet++ with Improved Training and Scaling Strategies
- Reinforcement Learning with Neural Radiance Fields

Motion
- Unsupervised 3D Object Segmentation from Rigid Dynamics of Point Clouds

- Unsupervised Multi-object Segmentation by Predicting Probable Motion Patterns

Imitation Learning / Reinforcement Learning
- Model-based Imitation Learning for Urban Driving

- Human-AI Shared Control via Policy Dissection

- Behavior Transformers: Cloning k modes with one stone
- Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline
- Reinforcement Learning with a Terminator
- Imitation is Not Enough: Imitation with Reinforcement Learning for Challenging Driving Scenarios
- Meta Drive: Composing Diverse Driving Scenarios for Generalizable Reinforcement Learning
- Traffic Gen: Learning to Generate Diverse and Realistic Traffic Scenarios

- Human-in-the-Loop Deep Reinforcement Learning with Application to Autonomous Driving

Dataset / Benchmark
- m RI: Multi-modal 3D Human Pose Estimation Dataset using mmWave

- xView3-SAR: Detecting Dark Fishing Activity

- Mind Reader: Reconstructing Complex Images from Brain Activities
- Safe Bench: A Benchmarking Platform for Safety Evaluation

발표자: 백동희 박사과정

일시:  2022년 12월 20일 (1회차), 12월 30일 (2회차)