SGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation

Keyphrases

• Transformer 100%
• 3D Instance Segmentation 100%
• Semantic-guided 100%
• Transformer Encoder 66%
• Geometric Properties 33%
• Initialization Method 33%
• Estimation Bias 33%
• Semantic Information 33%
• Transformer Model 33%
• Publicly Available 33%
• Point Coordinates 33%
• Art Performance 33%
• Complex Design 33%
• Stacked Layer 33%
• Point Cloud Segmentation 33%
• Global Scene 33%
• Query Set 33%
• Auxiliary Task 33%
• Initialization Problems 33%
• Scene-aware 33%
• Scene Feature 33%
• Voxel-wise 33%
• S3DIS 33%
• Large-scale 3D Scene 33%
• Code Weight 33%
• Scene Prior 33%

Computer Science

• Instance Segmentation 100%
• 3d Scenes 33%
• Point Cloud 33%
• Art Performance 33%
• Coordinate Point 33%
• Cloud Instance 33%