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

Enhancing knowledge distillation for semantic segmentation through text-assisted modular plugins

delete2025-05-01
delete0
PRE
AI
L
Letian Wu
S
Shen Zhang
C
Chuankai Zhang
Z
Zhenyu Zhao
J
Jiajun Liang
DOI：10.1016/j.patcog.2024.111329delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
与其他模型压缩方法（如剪枝和量化）相比，知识蒸馏提供了更优越的兼容性和灵活性。当前用于语义分割的知识蒸馏（KD）方法主要引导学生模型仅通过图像数据复制教师模型的结构化信息。然而，这些方法往往忽视了在蒸馏过程中引入辅助模态（如文本信息）的潜在益处，从而未能有效弥合学生模型与教师模型之间的差距。本文介绍了一种新颖的文本辅助蒸馏方法。利用对比语言-图像预训练（CLIP）框架，我们提出了两个模块化插件：文本通道蒸馏模块和文本区域蒸馏模块，旨在将文本先验整合到蒸馏过程中。这些模块作为学生模型与教师模型之间的桥梁，增强了学生模型对教师网络的模仿。这些模块的特点是简单、通用，并能与现有的知识蒸馏框架无缝集成，从而促进性能提升。在Cityscapes、Pascal VOC和CamVid数据集上的实验评估表明，将这些即插即用模块增强当前最先进蒸馏技术后，显著提高了蒸馏效果。
Keyword:
Semantic segmentation
Knowledge distillation
Modality fusion
Plug-and-play module

期刊

Pattern Recognition 封面图
Pattern Recognition
IF:
7.6
论文数:
1.3W
被引数:
4.5W

机构

暂无机构信息
引用论文

引用论文

MaskCLIP: Masked Self-Distillation Advances Contrastive Language-Image Pretraining
err2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2023-06-01
err0
errOAAI
errXiaoyi Dong; Jianmin Bao; Yinglin Zheng; Ting Zhang; Dongdong Chen; Hao Yang; Ming Zeng; Weiming Zhang; Lu Yuan; Dong Chen; Fang Wen; Nenghai Yu
err分享
err收藏
The Cityscapes Dataset for Semantic Urban Scene Understanding用于语义城市场景理解的Cityscapes数据集
err2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2016-06-01
err0
errOAAI
errMarius Cordts; Mohamed Omran; Sebastian Ramos; Timo Rehfeld; Markus Enzweiler; Rodrigo Benenson; Uwe Franke; Stefan Roth; Bernt Schiele
err分享
err收藏
Quantized Convolutional Neural Networks for Mobile Devices
err2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2016-06-01
err0
errOAAI
errJiaxiang Wu; Cong Leng; Yuhang Wang; Qinghao Hu; Jian Cheng
err分享
err收藏
Structured Knowledge Distillation for Dense Prediction
errIEEE Transactions on Pattern Analysis and Machine Intelligence
IF0
err2023-06-01
err0
errOAAI
errYifan Liu; Changyong Shu; Jingdong Wang; Chunhua Shen
err分享
err收藏
Channel-wise Knowledge Distillation for Dense Prediction
err2021 IEEE/CVF International Conference on Computer Vision (ICCV)
IF0
err2021-10-01
err0
errOAAI
errChangyong Shu; Yifan Liu; Jianfei Gao; Zheng Yan; Chunhua Shen
err分享
err收藏
DenseCLIP: Language-Guided Dense Prediction with Context-Aware Prompting
err2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2022-06-01
err0
errOAAI
errYongming Rao; Wenliang Zhao; Guangyi Chen; Yansong Tang; Zheng Zhu; Guan Huang; Jie Zhou; Jiwen Lu
err分享
err收藏
Adversarial co-distillation learning for image recognition
errPATTERN RECOGNITION
IF7.6
err2021-03-01
err28
PREAI
errZhang, Haoran; Hu, Zhenzhen; Qin, Wei; Xu, Mingliang; Wang, Meng
err分享
err收藏
Cross-Image Relational Knowledge Distillation for Semantic Segmentation
err2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2022-06-01
err0
errOAAI
errChuanguang Yang; Helong Zhou; Zhulin An; Xue Jiang; Yongjun Xu; Qian Zhang
err分享
err收藏
PointCLIP: Point Cloud Understanding by CLIP
err2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
IF0
err2022-06-01
err0
errOAAI
errRenrui Zhang; Ziyu Guo; Wei Zhang; Kunchang Li; Xupeng Miao; Bin Cui; Yu Qiao; Peng Gao; Hongsheng Li
err分享
err收藏
Explainable scale distillation for hyperspectral image classification用于高光谱图像分类的可解释尺度蒸馏
errPATTERN RECOGNITION
IF7.6
err2022-02-01
err28
PREAI
errShi, Cheng; Fang, Li; Lv, Zhiyong; Zhao, Minghua
err分享
err收藏
学者 查看更多内容