https://arxiv.org/pdf/2108.08367.pdf

Methodology

• Single layer method인 GRD-net 에서 3D reconstructure를 가져와서 2 layer로 구성하여 작동 시켰다고 하는데 자세한건 이해가 ㅇㅏㄴ간다
  
• Θ denoting the trainable parameters of the utilized network

Transformer 관련

• ViT AN IMAGE IS WORTH 16X16 WORDS:
  TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE https://arxiv.org/pdf/2010.11929.pdf
• DETR: End-to-End Object Detection with Transformers

Object detection

Object rotation

• DAFNe: A One-Stage Anchor-Free Deep Model for Oriented Object Detection https://arxiv.org/pdf/2109.06148v2.pdf
• ReDet: A Rotation-equivariant Detector for Aerial Object Detection https://arxiv.org/abs/2103.07733

Anchor free

• FCOS: Fully Convolutional One-Stage Object Detectionhttps://arxiv.org/pdf/1904.01355.pdf

NMS 보다 좋은 앙상블 방법 WBF

• Weighted boxes fusion: Ensembling boxes from different object detection modelshttps://arxiv.org/pdf/1910.13302.pdf

위성영상관련

• https://neverabandon.tistory.com/52

6dof pose estimation

• SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation
  https://arxiv.org/pdf/2108.08367.pdf#page=9&zoom=100,66,844
• Single-stage 6D Object Pose estimationhttps://arxiv.org/pdf/1911.08324.pdf
• Self6D: Self-Supervised Monocular 6D Object Pose Estimationhttps://arxiv.org/pdf/2004.06468.pdf
• EPOS: Estimating 6D Pose of Objects with Symmetrieshttps://arxiv.org/pdf/2004.00605.pdf

6dof grasp

• 6-DOF GraspNet: Variational Grasp Generation for Object Manipulation

3D reconstructure

• 3D Scene Reconstruction with Multi-layer Depth and Epipolar Transformershttps://arxiv.org/pdf/1902.06729.pdf
  -> SO-pose 부분에서 3D 생성하는 부분의 기반 논문

• dhkim0225/1day_1paper#15