Lightweight Instruction Set for Flexible Dilated Convolutions and Mixed-Precision Operands

Keyphrases

• Instruction Sets 100%
• Dilated Convolution 100%
• Mixed Precision 100%
• Operand 100%
• Computation Time 66%
• Dilation Rate 66%
• DeepLabv3+ 66%
• Transposed Convolution 66%
• Precision-scalable 66%
• Deep Neural Network 33%
• Object Detection 33%
• Semantic Segmentation 33%
• Feature Map 33%
• Hardware Overhead 33%
• Number of Zeros 33%
• Convolutional Layer 33%
• Limited Range 33%
• Object Semantics 33%
• Proposed Design 33%
• ResNet50 33%
• Standard Neural Network 33%
• ENet 33%
• Neural Network Accelerator 33%
• Instruction Memory 33%
• PointPillars 33%
• Redundant Elements 33%
• Memory Scheme 33%

Computer Science

• Computation Time 100%
• Speed-up 50%
• Image Segmentation 50%
• Neural Network 50%
• Object Detection 50%
• Deep Neural Network 50%
• Hardware Overhead 50%
• Residual Neural Network 50%
• Convolution Layer 50%
• Hardware Accelerator 50%
• Paying Attention 50%
• Instruction Memory 50%
• Computer Hardware 50%
• Feature Map 50%