A Systematic Performance Analysis of Deep Perceptual Loss Networks: Breaking Transfer Learning Conventions

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Pihlgren, Gustav Grund, Nikolaidou, Konstantina, Chhipa, Prakash Chandra, Abid, Nosheen, Saini, Rajkumar, Sandin, Fredrik, Liwicki, Marcus
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911941347246080
author Pihlgren, Gustav Grund
Nikolaidou, Konstantina
Chhipa, Prakash Chandra
Abid, Nosheen
Saini, Rajkumar
Sandin, Fredrik
Liwicki, Marcus
author_facet Pihlgren, Gustav Grund
Nikolaidou, Konstantina
Chhipa, Prakash Chandra
Abid, Nosheen
Saini, Rajkumar
Sandin, Fredrik
Liwicki, Marcus
contents In recent years, deep perceptual loss has been widely and successfully used to train machine learning models for many computer vision tasks, including image synthesis, segmentation, and autoencoding. Deep perceptual loss is a type of loss function for images that computes the error between two images as the distance between deep features extracted from a neural network. Most applications of the loss use pretrained networks called loss networks for deep feature extraction. However, despite increasingly widespread use, the effects of loss network implementation on the trained models have not been studied. This work rectifies this through a systematic evaluation of the effect of different pretrained loss networks on four different application areas. Specifically, the work evaluates 14 different pretrained architectures with four different feature extraction layers. The evaluation reveals that VGG networks without batch normalization have the best performance and that the choice of feature extraction layer is at least as important as the choice of architecture. The analysis also reveals that deep perceptual loss does not adhere to the transfer learning conventions that better ImageNet accuracy implies better downstream performance and that feature extraction from the later layers provides better performance.
format Preprint
id arxiv_https___arxiv_org_abs_2302_04032
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle A Systematic Performance Analysis of Deep Perceptual Loss Networks: Breaking Transfer Learning Conventions
Pihlgren, Gustav Grund
Nikolaidou, Konstantina
Chhipa, Prakash Chandra
Abid, Nosheen
Saini, Rajkumar
Sandin, Fredrik
Liwicki, Marcus
Computer Vision and Pattern Recognition
Machine Learning
In recent years, deep perceptual loss has been widely and successfully used to train machine learning models for many computer vision tasks, including image synthesis, segmentation, and autoencoding. Deep perceptual loss is a type of loss function for images that computes the error between two images as the distance between deep features extracted from a neural network. Most applications of the loss use pretrained networks called loss networks for deep feature extraction. However, despite increasingly widespread use, the effects of loss network implementation on the trained models have not been studied. This work rectifies this through a systematic evaluation of the effect of different pretrained loss networks on four different application areas. Specifically, the work evaluates 14 different pretrained architectures with four different feature extraction layers. The evaluation reveals that VGG networks without batch normalization have the best performance and that the choice of feature extraction layer is at least as important as the choice of architecture. The analysis also reveals that deep perceptual loss does not adhere to the transfer learning conventions that better ImageNet accuracy implies better downstream performance and that feature extraction from the later layers provides better performance.
title A Systematic Performance Analysis of Deep Perceptual Loss Networks: Breaking Transfer Learning Conventions
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2302.04032