3D Object Pose Estimation

On-going article

Recent works

• SSD-6D: Making RGB-Based 3D Detection and 6D Pose Estimation Great Again, ICCV2017
• PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes, RSS2018
• (RGB) Deep Object Pose Estimation for Semantic Robotic Grasping of Household Objects (DOPE), CoRL 2018
• (RGB) Real-time seamless Single Shot 6D Object Pose Prediction, CVPR2018
• (RGB+D) DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion, CVPR2019
• (RGB) DeepIM: Deep Iterative Matching for 6D Pose Estimation, ECCV2018
• (RGB+D) GRIP: Generative Robust Inference and Perception for Semantic Robot Manipulation in Adversarial Environments, IROS 2019

Metric

• Average distance (ADD) metric : average 3D distance of model vertices
  – 3D 공간상의 mean distance가 10% 안쪽이면 맞다고 함
  – ADD-S 라는 Closest Object에 대한 Metric도 있음
• 2D reproduction error (CVPR2018) 5Pixel 이하면 Correct 라고 하고 있음
• IoU Score (Projection 한 BB 0.5 이상이면)

Dataset

• YCB-Video
• LINEMOD Dataset

The goal of Our Project

• LineMOD Dataset
• DeepIM: Deep Iterative Matching for 6D Pose Estimation, ECCV2018
• ECCV2018 SOTA ADD 기준 88.6 
• 우리 목표 90, 92, 94 으로 잡으면 될 듯 싶음….