YolactEdge：边缘端实时实例分割
摘要
我们提出YolactEdge，这是首个可在小型边缘设备上以实时速度运行的具有竞争力的实例分割方法。具体而言，YolactEdge在Jetson AGX Xavier上以550x550分辨率图像、ResNet-101骨干网络可达30.8 FPS（在RTX 2080 Ti上达172.7 FPS）。为此，我们对最先进的基于图像的实时方法YOLACT做了两处改进：（1）应用TensorRT优化，同时谨慎权衡速度与精度；（2）一种新颖的特征扭曲模块，以利用视频中的时间冗余。在YouTube VIS和MS COCO数据集上的实验表明，YolactEdge相较现有实时方法实现了3-5倍加速，同时取得了具有竞争力的掩码与边界框检测精度。我们还进行了消融实验以剖析我们的设计选择与模块。代码与模型见https://github.com/haotian-liu/yolact_edge。
引用
@article{arxiv.2012.12259,
title = {YolactEdge: Real-time Instance Segmentation on the Edge},
author = {Haotian Liu and Rafael A. Rivera Soto and Fanyi Xiao and Yong Jae Lee},
journal= {arXiv preprint arXiv:2012.12259},
year = {2021}
}
备注
\c{opyright} 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works