An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures

The You Only Look Once (YOLO) architecture is crucial for real-time object detection. However, deploying it in resource-constrained environments such as unmanned aerial vehicles (UAVs) requires efficient transfer learning. Although layer freezing is a common technique, the specific impact of various...

Full description

Saved in:
Bibliographic Details
Main Authors: Andrzej D. Dobrzycki, Ana M. Bernardos, José R. Casar
Format: Article
Language:English
Published: MDPI AG 2025-08-01
Series:Mathematics
Subjects:
Online Access:https://www.mdpi.com/2227-7390/13/15/2539
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849765949634248704
author Andrzej D. Dobrzycki
Ana M. Bernardos
José R. Casar
author_facet Andrzej D. Dobrzycki
Ana M. Bernardos
José R. Casar
author_sort Andrzej D. Dobrzycki
collection DOAJ
description The You Only Look Once (YOLO) architecture is crucial for real-time object detection. However, deploying it in resource-constrained environments such as unmanned aerial vehicles (UAVs) requires efficient transfer learning. Although layer freezing is a common technique, the specific impact of various freezing configurations on contemporary YOLOv8 and YOLOv10 architectures remains unexplored, particularly with regard to the interplay between freezing depth, dataset characteristics, and training dynamics. This research addresses this gap by presenting a detailed analysis of layer-freezing strategies. We systematically investigate multiple freezing configurations across YOLOv8 and YOLOv10 variants using four challenging datasets that represent critical infrastructure monitoring. Our methodology integrates a gradient behavior analysis (L2 norm) and visual explanations (Grad-CAM) to provide deeper insights into training dynamics under different freezing strategies. Our results reveal that there is no universal optimal freezing strategy but, rather, one that depends on the properties of the data. For example, freezing the backbone is effective for preserving general-purpose features, while a shallower freeze is better suited to handling extreme class imbalance. These configurations reduce graphics processing unit (GPU) memory consumption by up to 28% compared to full fine-tuning and, in some cases, achieve mean average precision (mAP@50) scores that surpass those of full fine-tuning. Gradient analysis corroborates these findings, showing distinct convergence patterns for moderately frozen models. Ultimately, this work provides empirical findings and practical guidelines for selecting freezing strategies. It offers a practical, evidence-based approach to balanced transfer learning for object detection in scenarios with limited resources.
format Article
id doaj-art-6254331a91d94849a5cc3f578c92fb72
institution DOAJ
issn 2227-7390
language English
publishDate 2025-08-01
publisher MDPI AG
record_format Article
series Mathematics
spelling doaj-art-6254331a91d94849a5cc3f578c92fb722025-08-20T03:04:43ZengMDPI AGMathematics2227-73902025-08-011315253910.3390/math13152539An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO ArchitecturesAndrzej D. Dobrzycki0Ana M. Bernardos1José R. Casar2Information Processing and Telecommunications Center, Universidad Politécnica de Madrid, ETSI Telecomunicación, Av. Complutense, 30, 28040 Madrid, SpainInformation Processing and Telecommunications Center, Universidad Politécnica de Madrid, ETSI Telecomunicación, Av. Complutense, 30, 28040 Madrid, SpainInformation Processing and Telecommunications Center, Universidad Politécnica de Madrid, ETSI Telecomunicación, Av. Complutense, 30, 28040 Madrid, SpainThe You Only Look Once (YOLO) architecture is crucial for real-time object detection. However, deploying it in resource-constrained environments such as unmanned aerial vehicles (UAVs) requires efficient transfer learning. Although layer freezing is a common technique, the specific impact of various freezing configurations on contemporary YOLOv8 and YOLOv10 architectures remains unexplored, particularly with regard to the interplay between freezing depth, dataset characteristics, and training dynamics. This research addresses this gap by presenting a detailed analysis of layer-freezing strategies. We systematically investigate multiple freezing configurations across YOLOv8 and YOLOv10 variants using four challenging datasets that represent critical infrastructure monitoring. Our methodology integrates a gradient behavior analysis (L2 norm) and visual explanations (Grad-CAM) to provide deeper insights into training dynamics under different freezing strategies. Our results reveal that there is no universal optimal freezing strategy but, rather, one that depends on the properties of the data. For example, freezing the backbone is effective for preserving general-purpose features, while a shallower freeze is better suited to handling extreme class imbalance. These configurations reduce graphics processing unit (GPU) memory consumption by up to 28% compared to full fine-tuning and, in some cases, achieve mean average precision (mAP@50) scores that surpass those of full fine-tuning. Gradient analysis corroborates these findings, showing distinct convergence patterns for moderately frozen models. Ultimately, this work provides empirical findings and practical guidelines for selecting freezing strategies. It offers a practical, evidence-based approach to balanced transfer learning for object detection in scenarios with limited resources.https://www.mdpi.com/2227-7390/13/15/2539YOLOlayer freezingtransfer learningfine-tuningobject detection
spellingShingle Andrzej D. Dobrzycki
Ana M. Bernardos
José R. Casar
An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
Mathematics
YOLO
layer freezing
transfer learning
fine-tuning
object detection
title An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
title_full An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
title_fullStr An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
title_full_unstemmed An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
title_short An Analysis of Layer-Freezing Strategies for Enhanced Transfer Learning in YOLO Architectures
title_sort analysis of layer freezing strategies for enhanced transfer learning in yolo architectures
topic YOLO
layer freezing
transfer learning
fine-tuning
object detection
url https://www.mdpi.com/2227-7390/13/15/2539
work_keys_str_mv AT andrzejddobrzycki ananalysisoflayerfreezingstrategiesforenhancedtransferlearninginyoloarchitectures
AT anambernardos ananalysisoflayerfreezingstrategiesforenhancedtransferlearninginyoloarchitectures
AT josercasar ananalysisoflayerfreezingstrategiesforenhancedtransferlearninginyoloarchitectures
AT andrzejddobrzycki analysisoflayerfreezingstrategiesforenhancedtransferlearninginyoloarchitectures
AT anambernardos analysisoflayerfreezingstrategiesforenhancedtransferlearninginyoloarchitectures
AT josercasar analysisoflayerfreezingstrategiesforenhancedtransferlearninginyoloarchitectures