Weight quantization retraining for sparse and compressed spatial domain correlation filters

Computer Science

• Compression Ratio 36%
• Computational Resource 9%
• Data Type 9%
• Fixed Points 18%
• Geometric Transformation 9%
• Internet of Things 9%
• Particle Swarm Optimization 9%
• Pattern Recognition 36%
• Preprocessing Step 9%
• Quantisation Error 18%
• Quantization (Signal Processing) 100%
• Quantization Level 9%
• Quantization Noise 9%
• Sparsity 9%
• Spatial Domain 100%
• Training Sample 9%

Keyphrases

• Accuracy Degradation 50%
• Correlation Pattern Recognition 100%
• Dynamic Fixed Point 50%
• Log-polar 50%
• Log-polar Transform 50%
• Power-of-two Quantization 50%
• Trained Filter 25%
• Weight Compression 25%

Engineering

• Geometric Transformation 9%