2D-3D Interlaced Transformer for Point Cloud Segmentation with Scene-Level Supervision

Keyphrases

• Transformer 100%
• 2D to 3D 100%
• Point Cloud Segmentation 100%
• Unsupervised Learning 80%
• 2D Features 80%
• 3D Features 80%
• Decoder 60%
• Encoder 40%
• Feature Fusion 40%
• 3D Point Cloud 20%
• Segmentation Method 20%
• Multi-view Images 20%
• Point Cloud 20%
• 3D Information 20%
• Information Fusion 20%
• Large Margin 20%
• Annotation Cost 20%
• 3D Data 20%
• 2D Data 20%
• Transformer Model 20%
• S3DIS 20%
• Cross-attention 20%
• Key-value Pairs 20%

Computer Science

• Point Cloud Segmentation 100%
• Annotation 40%
• Point Cloud 40%
• Feature Fusion 40%
• Coder 40%
• Supervised Learning 20%
• Segmentation Method 20%

Engineering

• Point Cloud 100%
• Segmentation Method 14%