CARD: Semantic Segmentation with Efficient Class-Aware Regularized Decoder

Keyphrases

• Decoder 100%
• Semantic Segmentation 100%
• Class-aware Regularizations 100%
• Class Center 80%
• Class Level 60%
• High Efficiency 40%
• Contextual Information 40%
• Object-contextual Representations 40%
• Segmentation Model 20%
• Concatenated 20%
• Loss Function 20%
• Computational Cost 20%
• Benchmark Dataset 20%
• Regularization Method 20%
• Generalization Ability 20%
• Token 20%
• Error-prone 20%
• Feature Learning 20%
• Error Accumulation 20%
• Ablation Study 20%
• Context Prior 20%
• Representation Learning 20%
• Inter-class Distance 20%
• Intra-class Variation 20%
• Class Representation 20%
• Universal Classes 20%
• Mask Estimation 20%
• First-loss 20%
• Coarse Prediction 20%
• Pixel Features 20%
• Novel Loss Function 20%
• Center-based 20%
• Upsampling 20%
• Mixer 20%
• Efficient Architecture 20%

Computer Science

• Upsampling 20%