CustomKD: Customizing Large Vision Foundation for Edge Model Improvement via Knowledge Distillation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lee, Jungsoo, Das, Debasmit, Hayat, Munawar, Choi, Sungha, Hwang, Kyuwoong, Porikli, Fatih
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913754020577280
author Lee, Jungsoo
Das, Debasmit
Hayat, Munawar
Choi, Sungha
Hwang, Kyuwoong
Porikli, Fatih
author_facet Lee, Jungsoo
Das, Debasmit
Hayat, Munawar
Choi, Sungha
Hwang, Kyuwoong
Porikli, Fatih
contents We propose a novel knowledge distillation approach, CustomKD, that effectively leverages large vision foundation models (LVFMs) to enhance the performance of edge models (e.g., MobileNetV3). Despite recent advancements in LVFMs, such as DINOv2 and CLIP, their potential in knowledge distillation for enhancing edge models remains underexplored. While knowledge distillation is a promising approach for improving the performance of edge models, the discrepancy in model capacities and heterogeneous architectures between LVFMs and edge models poses a significant challenge. Our observation indicates that although utilizing larger backbones (e.g., ViT-S to ViT-L) in teacher models improves their downstream task performances, the knowledge distillation from the large teacher models fails to bring as much performance gain for student models as for teacher models due to the large model discrepancy. Our simple yet effective CustomKD customizes the well-generalized features inherent in LVFMs to a given student model in order to reduce model discrepancies. Specifically, beyond providing well-generalized original knowledge from teachers, CustomKD aligns the features of teachers to those of students, making it easy for students to understand and overcome the large model discrepancy overall. CustomKD significantly improves the performances of edge models in scenarios with unlabeled data such as unsupervised domain adaptation (e.g., OfficeHome and DomainNet) and semi-supervised learning (e.g., CIFAR-100 with 400 labeled samples and ImageNet with 1% labeled samples), achieving the new state-of-the-art performances.
format Preprint
id arxiv_https___arxiv_org_abs_2503_18244
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CustomKD: Customizing Large Vision Foundation for Edge Model Improvement via Knowledge Distillation
Lee, Jungsoo
Das, Debasmit
Hayat, Munawar
Choi, Sungha
Hwang, Kyuwoong
Porikli, Fatih
Computer Vision and Pattern Recognition
We propose a novel knowledge distillation approach, CustomKD, that effectively leverages large vision foundation models (LVFMs) to enhance the performance of edge models (e.g., MobileNetV3). Despite recent advancements in LVFMs, such as DINOv2 and CLIP, their potential in knowledge distillation for enhancing edge models remains underexplored. While knowledge distillation is a promising approach for improving the performance of edge models, the discrepancy in model capacities and heterogeneous architectures between LVFMs and edge models poses a significant challenge. Our observation indicates that although utilizing larger backbones (e.g., ViT-S to ViT-L) in teacher models improves their downstream task performances, the knowledge distillation from the large teacher models fails to bring as much performance gain for student models as for teacher models due to the large model discrepancy. Our simple yet effective CustomKD customizes the well-generalized features inherent in LVFMs to a given student model in order to reduce model discrepancies. Specifically, beyond providing well-generalized original knowledge from teachers, CustomKD aligns the features of teachers to those of students, making it easy for students to understand and overcome the large model discrepancy overall. CustomKD significantly improves the performances of edge models in scenarios with unlabeled data such as unsupervised domain adaptation (e.g., OfficeHome and DomainNet) and semi-supervised learning (e.g., CIFAR-100 with 400 labeled samples and ImageNet with 1% labeled samples), achieving the new state-of-the-art performances.
title CustomKD: Customizing Large Vision Foundation for Edge Model Improvement via Knowledge Distillation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.18244