Convolutional Neural Network Model Compression Method for Software—Hardware Co-Design

Round 1

Reviewer 1 Report

The authors developed an automation technique by combining model compression for accelerating CNN models and methods using CPU–FPGA parallelism. The experimental results demonstrated that our design is more speed up than the conventional implementation method. Moreover, we evaluated this method on the ZCU FPGA platforms, and it achieved 2.47x, 1.93x, and 2.16x speed up for MobileNetV1, ShuffleNetV2, and SqueezeNet, respectively.  I highly encourage authors to deposit their code in public repository like Github, gitlab, etc. to ensure reproduction of results.

Author Response

Responses to Reviewer 1

We would like to thank you and the review team for seeing the potential in our paper and giving us constructive guidelines for revision. We have revised the paper thoroughly based on the comments and feedback we received. We provide a summary of the major improvements to the manuscript; we then respond to individual comments by each member of the review team.

Comment:

I highly encourage authors to deposit their code in public repository like Github, gitlab, etc. to ensure reproduction of results.

Response:

Thank you for your advice. 

We will complete the project at <https://gitlab.com/seojinygud/cnn-compression-with-cpu-fpga-co-design>. 

However, since it is currently research conducted with project support, it is impossible to disclose it, so we plan to disclose it after the completion of the project. The above site will pass it on to the editor.

Reviewer 2 Report

Author Response

We would like to thank you and the review team for seeing the potential in our paper and giving us constructive guidelines for revision.