Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Aug 13, 2025

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.3K

A Hardware-Friendly High-Precision CNN Pruning Method and Its FPGA Implementation.

Xuefu Sui1,2,3, Qunbo Lv1,2,3, Liangjie Zhi1,2,3

  • 1Aerospace Information Research Institute, Chinese Academy of Sciences, No. 9 Dengzhuang South Road, Haidian District, Beijing 100094, China.

Sensors (Basel, Switzerland)
|January 21, 2023
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Dynamic Evolution-Controlled Parabolic-Shaped Microstructures for Ultra-Black Surface via Self-Assembled Microsphere Mask Etching.

Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
Same author

Infrared Small-Target Segmentation Framework Based on Morphological Attention and Energy Core Loss.

Journal of imaging·2026
Same author

WeatherMAR: Complementary Masking of Paired Tokens for Adverse-Weather Image Restoration.

Journal of imaging·2026
Same author

Case Report: One case of pregnancy complicated by large cell neuroendocrine carcinoma of the cervix with syndrome of inappropriate secretion of antidiuretic hormone.

Frontiers in oncology·2025
Same author

TREM2 deficiency exacerbates cognitive impairment by aggravating α-Synuclein-induced lysosomal dysfunction in Parkinson's disease.

Cell death discovery·2025
Same author

Spectral Super-Resolution Technology Based on Fabry-Perot Interferometer for Temporally and Spatially Modulated Fourier Transform Imaging Spectrometer.

Sensors (Basel, Switzerland)·2025
This summary is machine-generated.

We developed KRP, a novel convolutional neural network (CNN) pruning method that reduces storage and improves efficiency for hardware deployment. This method achieves high accuracy and significant model compression, benefiting edge computing applications.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Hardware Acceleration

Background:

  • Convolutional Neural Networks (CNNs) face challenges in hardware deployment due to large storage needs and computational demands.
  • Efficient deployment of CNNs is crucial for edge computing applications, requiring optimized model parameters and hardware utilization.

Purpose of the Study:

  • To develop an innovative, hardware-friendly CNN pruning method to address storage and efficiency issues.
  • To create a CNN compression framework that achieves high pruning rates and accuracy for practical hardware deployment.

Main Methods:

  • Introduced KRP, a novel CNN pruning method that operates on a convolutional kernel row scale.
  • Employed a retraining strategy based on LR tracking to maintain high accuracy post-pruning.
Keywords:
LR trackingconvolutional neural networkshardware friendlyhigh parallelismnetwork compressionregular pruning

More Related Videos

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
04:23

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

1.9K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

600

Related Experiment Videos

Last Updated: Aug 13, 2025

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.3K
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
04:23

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images

Published on: April 21, 2023

1.9K
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

600
  • Designed a high-performance convolutional computation module for FPGA deployment of pruned models.

    • Main Results:

    • KRP demonstrated superior accuracy compared to existing pruning methods on benchmark CNNs like VGG and ResNet.
    • The KRP and GSNQ quantization framework achieved "lossless" CNN compression, reducing model storage by 27×.
    • FPGA experiments confirmed KRP's significant reduction in storage and on-chip resource consumption, alongside improved parallelism.

    Conclusions:

    • KRP offers a hardware-friendly solution for efficient CNN compression, crucial for edge computing.
    • The developed framework provides a pathway for deploying high-performance, compressed CNNs on resource-constrained hardware.
    • This research contributes novel insights into optimizing CNNs for edge AI through effective pruning and hardware co-design.