GnetDet: Object Detection Optimized on a 224mW CNN Accelerator Chip at the Speed of 106FPS

TL;DR

GnetDet is an optimized object detection model designed for CNN accelerator chips, achieving 106FPS at low power consumption of 224mW with minimal CPU load and high accuracy.

Contribution

The paper introduces GnetDet, a new object detection model optimized specifically for CNN accelerator chips to maximize speed and minimize CPU load.

Findings

Achieves 106FPS on a 224mW CNN accelerator chip.

Maintains high accuracy with low power consumption.

Reduces CPU load during inference.

Abstract

Object detection is widely used on embedded devices. With the wide availability of CNN (Convolutional Neural Networks) accelerator chips, the object detection applications are expected to run with low power consumption, and high inference speed. In addition, the CPU load is expected to be as low as possible for a CNN accelerator chip working as a co-processor with a host CPU. In this paper, we optimize the object detection model on the CNN accelerator chip by minimizing the CPU load. The resulting model is called GnetDet. The experimental result shows that the GnetDet model running on a 224mW chip achieves the speed of 106FPS with excellent accuracy.