Skip to content

Latest commit

 

History

History
43 lines (29 loc) · 1.37 KB

README.md

File metadata and controls

43 lines (29 loc) · 1.37 KB
Raw

VDDIB-SR

This repository aims at introducing the VDDIB-SR method, which can effectively reduce the communication overhead in multi-device edge inference. This method is proposed in our paper "Task-Oriented Communication for Multi-Device Cooperative Edge Inference", which is accepted to IEEE Transaction on Wireless Communication.

We release the code to reproduce some experimental results. Particularly, we focus on the multi-view image classification task on the MNIST dataset and the multi-view object recognition task on the ModelNet40 dataset. The codes for these two tasks are in the folder ./Multi-view_Image_Classification and folder ./Multi-view_Object_Recognition, respectively.

Dependencies

Packages

Pytorch 1.11.0
Torchvision 0.12.0
scikit-image 0.19.3

Datasets

MNIST
ModelNet40

Citation

@ARTICLE{9837474,
  author={Shao, Jiawei and Mao, Yuyi and Zhang, Jun},
  journal={IEEE Transactions on Wireless Communications}, 
  title={Task-Oriented Communication for Multidevice Cooperative Edge Inference}, 
  year={2023},
  volume={22},
  number={1},
  pages={73-87},
  doi={10.1109/TWC.2022.3191118}}

Others

  • (2023-Sep-12) I have corrected some typos and updated the paper on arxiv.