Network Architecture

Detail이 명시된 것들

• Layer 구성
• Input dimension
• Dilation ratio

Detail이 명시되지 않은 것들

• C param → Experiment
• Dropout ratio → Experiment
• 3D downsampling strategy → 구현 당시에 찾기 (3D model)

Training Strategy

Detail이 명시된 것들

• P-Net initial training sequence

Detail이 명시되지 않은 것들

• PairwiseNet (Training sample, Structure) → CRF-RNN implementation 보기

Geodesic Distance

Detail이 명시된 것들

• Formulation
• Implementation (Official)

Detail이 명시되지 않은 것들

• Weight (Euc-dist, Geo-dist) → Minor (우선은 Geo : 1)