Memory-efficient deep learning inference with incremental weight loading and data layout reorganization on edge systems

Keyphrases

• Memory Efficiency 100%
• System Application 100%
• Data Layout 100%
• Weighted Data 100%
• Edge System 100%
• Weight Loading 100%
• Deep Learning Inference 100%
• Model Inference 100%
• Efficient Deep Learning 100%
• Management Effectiveness 50%
• Embedded Devices 50%
• Memory Management 50%
• Memory Space 50%
• Quality of Service 50%
• Computing Power 50%
• Inference Model 50%
• Memory Consumption 50%
• Deep Neural Network 50%
• In-memory Computing 50%
• Network Delay 50%
• Leakage Problem 50%
• Edge Devices 50%
• Time Overhead 50%
• Image Classification 50%
• Face Recognition 50%
• Pattern Recognition Applications 50%
• Spatial Dimension 50%
• Memory Limitations 50%
• Power Limit 50%
• Privacy Leakage 50%
• Product Detection 50%
• Incremental Loading 50%
• Inference Time 50%
• Agricultural Products 50%
• Model Weight 50%
• Loading Strategy 50%
• Deep Neural Network Inference 50%

Computer Science

• Inference Model 100%
• Data Layout 100%
• Deep Learning Method 100%
• Application System 66%
• Deep Neural Network 66%
• Experimental Result 33%
• Embedded Device 33%
• Quality of Service 33%
• Memory Management 33%
• Computing Power 33%
• Network Inference 33%
• Face Recognition 33%
• Memory Consumption 33%
• Image Classification 33%
• Spatial Dimension 33%
• Privacy Leakage 33%
• Pattern Recognition 33%

Engineering

• Deep Learning Method 100%
• Deep Neural Network 100%
• Experimental Result 50%
• Pattern Recognition 50%
• Limitations 50%
• Quality of Service 50%
• Computing Power 50%
• Memory Space 50%
• Image Classification 50%
• Spatial Dimension 50%
• Improving Device 50%
• Embedded Device 50%