logo
科言猫学术研究的AI总结
language
手机端手机端
  • Phone Feature微信小程序
简体中文简体中文
notice消息通知
博客博客
delete
arrow
返回

PDNet by Partial Deep Convolution: A Better Lightweight Detector

delete2025-02-02
delete0
delete
OA
AI
W
Wei Wang
Y
Yuanze Meng *
H
Han Li
S
Shun Li
C
Chenghong Zhang
张光辉 封面图
张光辉 (Guanghui Zhang)
W
Weimin Lei *
DOI：10.3390/electronics14030591delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
模型轻量化在边缘计算和移动设备中具有重要意义。当前关于快速网络设计的研究主要集中于模型计算压缩和加速。许多模型旨在通过处理冗余特征图来压缩模型。然而，这些方法大多选择通过简单操作保留特征图，并未有效减少冗余特征图。本文提出了一种新的卷积模块PDConv，通过压缩冗余特征图来降低网络复杂度，并通过增加网络宽度来保持精度。PDConv（Partial Deep Convolution）在处理冗余特征图方面优于传统方法，尤其在深度网络中表现突出。其FLOPs与深度可分离卷积相当，但精度更高。本文提出了PDBottleNeck和PDC2f（Partial Deep CSPDarknet53 to 2-Stage FPN），并构建了轻量级网络PDNet，使用PASCAL VOC数据集进行实验验证。与流行的HorNet相比，本方法在FLOPs上实现了超过25%的提升，在mAP50:95精度上提升了1.8%。在CoCo2017数据集上，本方法的大规模PDNet在mAP75上实现了0.5%的提升，且FLOPs低于最新的RepVit。
Keyword:
PDConv
PDNet
lightweight
object detection

期刊

Electronics 封面图
Electronics
IF:
2.6
论文数:
1.0W
被引数:
4.7W

机构

Z
Zhejiang Univ Technol
学者数:
2.7K
论文数: 978
被引数: 382
N
Northeastern Univ
学者数:
2.9K
论文数: 1.3K
被引数: 362
引用论文

引用论文

FBNetV2: Differentiable Neural Architecture Search for Spatial and Channel Dimensions
err2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2020-06-01
err0
errOAAI
errAlvin Wan; Xiaoliang Dai; Peizhao Zhang; Zijian He; Yuandong Tian; Saining Xie; Bichen Wu; Matthew Yu; Tao Xu; Kan Chen; Peter Vajda; Joseph E. Gonzalez
err分享
err收藏
FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search
err2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2019-06-01
err0
errOAAI
errBichen Wu; Kurt Keutzer; Xiaoliang Dai; Peizhao Zhang; Yanghan Wang; Fei Sun; Yiming Wu; Yuandong Tian; Peter Vajda; Yangqing Jia
err分享
err收藏
ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile DevicesShuffleNet: 一种用于移动设备的极其高效的卷积神经网络
err2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition
IF0
err2018-06-01
err0
errOAAI
errXiangyu Zhang; Xinyu Zhou; Mengxiao Lin; Jian Sun
err分享
err收藏
CBAM: Convolutional Block Attention ModuleCBAM: 卷积块注意力模块
errLecture Notes in Computer Science
IF0
err2018-10-06
err0
PREAI
errSanghyun Woo; Jongchan Park; Joon-Young Lee; In So Kweon
err分享
err收藏
GhostNet: More Features From Cheap OperationsGhostNet: 廉价运营的更多功能
err2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2020-06-01
err0
errOAAI
errKai Han; Yunhe Wang; Qi Tian; Jianyuan Guo; Chunjing Xu; Chang Xu
err分享
err收藏
ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture DesignShuffleNet V2: 高效CNN架构设计的实用指南
errLecture Notes in Computer Science
IF0
err2018-10-09
err0
PREAI
errNingning Ma; Xiangyu Zhang; Hai-Tao Zheng; Jian Sun
err分享
err收藏
CondenseNet V2: Sparse Feature Reactivation for Deep Networks
err2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2021-06-01
err0
errOAAI
errLe Yang; Haojun Jiang; Ruojin Cai; Yulin Wang; Shiji Song; Gao Huang; Qi Tian
err分享
err收藏
ImageNet Classification with Deep Convolutional Neural Networks基于深度卷积神经网络的ImageNet分类
errCOMMUNICATIONS OF THE ACM
IF12.2
err2017-05-24
err8.3W
errOAAI
errKrizhevsky, Alex; Sutskever, Ilya; Hinton, Geoffrey E.
err分享
err收藏
CSPNet: A New Backbone that can Enhance Learning Capability of CNN
err2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
IF0
err2020-06-01
err0
errOAAI
errChien-Yao Wang; Hong-Yuan Mark Liao; Yueh-Hua Wu; Ping-Yang Chen; Jun-Wei Hsieh; I-Hau Yeh
err分享
err收藏
Rep ViT: Revisiting Mobile CNN From ViT PerspectiveRep ViT: 从ViT角度重新审视移动CNN
err2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2024-06-16
err0
PREAI
errAo Wang; Hui Chen; Zijia Lin; Jungong Han; Guiguang Ding
err分享
err收藏
学者 查看更多内容