Saved in:
Bibliographic Details
Main Authors: Yan, Zhaoyi, Liu, Kangjun, Ye, Qixiang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.21269
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909555075579904
author Yan, Zhaoyi
Liu, Kangjun
Ye, Qixiang
author_facet Yan, Zhaoyi
Liu, Kangjun
Ye, Qixiang
contents Knowledge distillation has become a cornerstone technique in deep learning, facilitating the transfer of knowledge from complex models to lightweight counterparts. Traditional distillation approaches focus on transferring knowledge at the instance level, but fail to capture nuanced semantic relationships within the data. In response, this paper introduces a novel methodology, Semantics-based Relation Knowledge Distillation (SeRKD), which reimagines knowledge distillation through a semantics-relation lens among each sample. By leveraging semantic components, \ie, superpixels, SeRKD enables a more comprehensive and context-aware transfer of knowledge, which skillfully integrates superpixel-based semantic extraction with relation-based knowledge distillation for a sophisticated model compression and distillation. Particularly, the proposed method is naturally relevant in the domain of Vision Transformers (ViTs), where visual tokens serve as fundamental units of representation. Experimental evaluations on benchmark datasets demonstrate the superiority of SeRKD over existing methods, underscoring its efficacy in enhancing model performance and generalization capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2503_21269
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Delving Deep into Semantic Relation Distillation
Yan, Zhaoyi
Liu, Kangjun
Ye, Qixiang
Computer Vision and Pattern Recognition
Knowledge distillation has become a cornerstone technique in deep learning, facilitating the transfer of knowledge from complex models to lightweight counterparts. Traditional distillation approaches focus on transferring knowledge at the instance level, but fail to capture nuanced semantic relationships within the data. In response, this paper introduces a novel methodology, Semantics-based Relation Knowledge Distillation (SeRKD), which reimagines knowledge distillation through a semantics-relation lens among each sample. By leveraging semantic components, \ie, superpixels, SeRKD enables a more comprehensive and context-aware transfer of knowledge, which skillfully integrates superpixel-based semantic extraction with relation-based knowledge distillation for a sophisticated model compression and distillation. Particularly, the proposed method is naturally relevant in the domain of Vision Transformers (ViTs), where visual tokens serve as fundamental units of representation. Experimental evaluations on benchmark datasets demonstrate the superiority of SeRKD over existing methods, underscoring its efficacy in enhancing model performance and generalization capabilities.
title Delving Deep into Semantic Relation Distillation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.21269