YOLO-LITE: A Real-Time Object Detection Algorithm Optimized for Non-GPU Computers

TL;DR

YOLO-LITE is a lightweight, efficient real-time object detection model optimized for non-GPU devices, achieving high speed and reasonable accuracy, making real-time detection accessible on portable hardware.

Contribution

The paper introduces YOLO-LITE, a smaller and faster object detection model based on YOLOV2, optimized for devices without GPUs, with significant speed improvements over existing models.

Findings

Achieved 33.81% mAP on PASCAL VOC dataset.

Runs at 21 FPS on non-GPU hardware.

Outperforms SSD MobilenetvI by 3.8x in speed.

Abstract

This paper focuses on YOLO-LITE, a real-time object detection model developed to run on portable devices such as a laptop or cellphone lacking a Graphics Processing Unit (GPU). The model was first trained on the PASCAL VOC dataset then on the COCO dataset, achieving a mAP of 33.81% and 12.26% respectively. YOLO-LITE runs at about 21 FPS on a non-GPU computer and 10 FPS after implemented onto a website with only 7 layers and 482 million FLOPS. This speed is 3.8x faster than the fastest state of art model, SSD MobilenetvI. Based on the original object detection algorithm YOLOV2, YOLO- LITE was designed to create a smaller, faster, and more efficient model increasing the accessibility of real-time object detection to a variety of devices.