ForkNet: Multi-branch volumetric semantic completion from a single depth image

Keyphrases

• Latent Space 100%
• Multi-branch 100%
• Semantic Scene Completion 100%
• Single Depth Image 100%
• Real Scene 100%
• Training Data 50%
• 3D Objects 50%
• Semantic Representation 50%
• Occlusion 50%
• Training Samples 50%
• Existing Data 50%
• Image-based 50%
• Encoder 50%
• Amount of Training 50%
• New Dataset 50%
• Network Layer 50%
• Geometric Representation 50%
• Image Completion 50%
• Completion Task 50%
• 3D Scene 50%
• Semantic 3D 50%
• Object Completion 50%
• 3D Semantics 50%
• Semantic Surfaces 50%
• Task Semantics 50%
• Real Noise 50%
• Semantic Branch 50%
• Multiple Discriminators 50%

Computer Science

• Completion Semantics 100%
• 3d Scenes 50%
• Discriminator 50%
• Transfer Information 50%
• Network Layer 50%
• Task Completion 50%
• Realism 50%
• Training Sample 50%
• Semantic Representation 50%
• Geometric Representation 50%
• Training Dataset 50%