CYFLOD: Cyclic Filtering and Loss Damping for Alleviating Noisy Labels in Fine-Grained Visual Classification

Keyphrases

• Noisy Labels 100%
• Fine-grained Visual Categorization 100%
• Dual-stage 66%
• Learning with Noisy Labels 66%
• Loss Function 33%
• Learning Framework 33%
• Classification Task 33%
• Data Quality 33%
• Transfer Learning 33%
• Staged Approach 33%
• Diverse Data 33%
• Labeling Method 33%
• Fine-grained Classification 33%
• CIFAR-10 33%
• Integration Scheme 33%
• Filtering Scheme 33%
• Model Generalization 33%
• Noise Model 33%
• Iterative Filtering 33%
• Standard Transfer 33%
• Class Overlap 33%
• Symmetry Condition 33%
• Real-world Noise 33%
• Conventional Learning 33%
• Asymmetric Condition 33%

Engineering

• Classification Task 100%
• Quality Data 100%
• Transfer Learning 100%
• Loss Function 100%
• Shortfall 100%
• Solution Stage 100%

Computer Science

• Learning Framework 100%
• Classification Task 100%
• Solution Stage 100%
• Transfer Learning 100%