Domain-Aware Fine-Tuning of Foundation Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Kaplan, Ugur Ali, Keuper, Margret, Khoreva, Anna, Zhang, Dan, Li, Yumeng
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910521131794432
author Kaplan, Ugur Ali
Keuper, Margret
Khoreva, Anna
Zhang, Dan
Li, Yumeng
author_facet Kaplan, Ugur Ali
Keuper, Margret
Khoreva, Anna
Zhang, Dan
Li, Yumeng
contents Foundation models (FMs) have revolutionized computer vision, enabling effective learning across different domains. However, their performance under domain shift is yet underexplored. This paper investigates the zero-shot domain adaptation potential of FMs by comparing different backbone architectures and introducing novel domain-aware components that leverage domain related textual embeddings. We propose domain adaptive normalization, termed as Domino, which explicitly leverages domain embeddings during fine-tuning, thus making the model domain aware. Ultimately, Domino enables more robust computer vision models that can adapt effectively to various unseen domains.
format Preprint
id arxiv_https___arxiv_org_abs_2407_03482
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Domain-Aware Fine-Tuning of Foundation Models
Kaplan, Ugur Ali
Keuper, Margret
Khoreva, Anna
Zhang, Dan
Li, Yumeng
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Foundation models (FMs) have revolutionized computer vision, enabling effective learning across different domains. However, their performance under domain shift is yet underexplored. This paper investigates the zero-shot domain adaptation potential of FMs by comparing different backbone architectures and introducing novel domain-aware components that leverage domain related textual embeddings. We propose domain adaptive normalization, termed as Domino, which explicitly leverages domain embeddings during fine-tuning, thus making the model domain aware. Ultimately, Domino enables more robust computer vision models that can adapt effectively to various unseen domains.
title Domain-Aware Fine-Tuning of Foundation Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2407.03482