CAR: Class-Aware Regularizations for Semantic Segmentation

Keyphrases

• Class Center 100%
• Class-aware Regularizations 100%
• Class Level 75%
• Pixel Features 50%
• Segmentation Method 25%
• Segmentation Model 25%
• Concatenated 25%
• Loss Function 25%
• Benchmark Dataset 25%
• Regularization Method 25%
• Generalization Ability 25%
• Error-prone 25%
• Feature Learning 25%
• Error Accumulation 25%
• Ablation Study 25%
• Network Modules 25%
• Representation Learning 25%
• Inter-class Distance 25%
• Intra-class Variation 25%
• Class Representation 25%
• Universal Classes 25%
• Use Classes 25%
• Mask Estimation 25%
• Complete Code 25%
• First-loss 25%
• Coarse Prediction 25%
• Novel Loss Function 25%
• Center-based 25%

Computer Science

• Regularization 100%
• Representation Learning 50%
• Optical Character Recognition 50%
• Segmentation Method 25%
• Segmentation Model 25%
• Baseline Model 25%
• Generalization Ability 25%
• Class Variance 25%