Feature-Balanced Loss for Long-Tailed Visual Recognition

Computer Science

• Training Data 100%
• Feature Space 100%
• Performance Degradation 100%
• Majority Class 100%
• Superior Performance 100%
• Deep Neural Network 100%
• Performance Gain 100%
• Long-Tailed Visual Recognition 100%
• Stimulus Intensity 100%

Keyphrases

• Balanced Loss Function 100%
• Long-tailed Recognition 100%
• Feature Norms 66%
• Training Data 33%
• State-of-the-art Techniques 33%
• Popular 33%
• Superior Performance 33%
• Performance Degradation 33%
• Feature Space 33%
• Majority Class 33%
• Deep Neural Network 33%
• Stimulus Intensity 33%
• Performance Gain 33%
• Biased Models 33%
• Long Tail 33%
• Curriculum Learning 33%
• Long-tail Problem 33%
• Recognition Benchmark 33%

Engineering

• State-of-the-Art Method 100%
• Performance Degradation 100%
• Space Data 100%
• Deep Neural Network 100%
• Feature Space 100%
• Predicted Result 100%
• Stimulus Intensity 100%