Multi-level transitional contrast learning for personalized image aesthetics assessment
Personalized image aesthetics assessment (PIAA) is aimed at modeling the unique aesthetic preferences of individuals, based on which personalized aesthetic scores are predicted. People have different standards for image aesthetics, and accordingly, images rated at the same aesthetic level by differe...
Saved in:
Main Authors: | , , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2023
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/171829 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-171829 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1718292023-11-09T04:02:33Z Multi-level transitional contrast learning for personalized image aesthetics assessment Yang, Zhichao Li, Leida Yang, Yuzhe Li, Yaqian Lin, Weisi School of Computer Science and Engineering Engineering::Computer science and engineering Contrast Learning Aesthetic Preferences Personalized image aesthetics assessment (PIAA) is aimed at modeling the unique aesthetic preferences of individuals, based on which personalized aesthetic scores are predicted. People have different standards for image aesthetics, and accordingly, images rated at the same aesthetic level by different users explicitly reveal their aesthetic preferences. However, previous PIAA models treat each individual as an isolated optimization target, failing to take full advantage of the contrastive information among users. Further, although people's aesthetic preferences are unique, they still share some commonalities, meaning that PIAA models could be built on the basis of generic aesthetics. Motivated by the above facts, this paper presents a Multi-level Transitional Contrast Learning (MTCL) framework for PIAA by transiting features from generic aesthetics to personalized aesthetics via contrastive learning. First, a generic image aesthetics assessment network is pre-trained to learn the common aesthetic features. Then, image sets rated to have the same aesthetic levels by different users are employed to learn the differentiated aesthetic features through multiple level-wise contrast learning based on the generic aesthetic features. Finally, a target user's PIAA model is built by integrating generic and differentiated aesthetic features. Extensive experiments on four benchmark PIAA databases demonstrate that the proposed MTCL model outperforms the state-of-the-arts. This work was supported in part by the National Natural Science Foundation of China under Grants 62171340, 61991451 and 61771473, and the OPPO Research Fund. 2023-11-09T04:02:33Z 2023-11-09T04:02:33Z 2023 Journal Article Yang, Z., Li, L., Yang, Y., Li, Y. & Lin, W. (2023). Multi-level transitional contrast learning for personalized image aesthetics assessment. IEEE Transactions On Multimedia. https://dx.doi.org/10.1109/TMM.2023.3290479 1520-9210 https://hdl.handle.net/10356/171829 10.1109/TMM.2023.3290479 2-s2.0-85163780632 en IEEE Transactions on Multimedia © 2023 IEEE. All rights reserved. |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Computer science and engineering Contrast Learning Aesthetic Preferences |
spellingShingle |
Engineering::Computer science and engineering Contrast Learning Aesthetic Preferences Yang, Zhichao Li, Leida Yang, Yuzhe Li, Yaqian Lin, Weisi Multi-level transitional contrast learning for personalized image aesthetics assessment |
description |
Personalized image aesthetics assessment (PIAA) is aimed at modeling the unique aesthetic preferences of individuals, based on which personalized aesthetic scores are predicted. People have different standards for image aesthetics, and accordingly, images rated at the same aesthetic level by different users explicitly reveal their aesthetic preferences. However, previous PIAA models treat each individual as an isolated optimization target, failing to take full advantage of the contrastive information among users. Further, although people's aesthetic preferences are unique, they still share some commonalities, meaning that PIAA models could be built on the basis of generic aesthetics. Motivated by the above facts, this paper presents a Multi-level Transitional Contrast Learning (MTCL) framework for PIAA by transiting features from generic aesthetics to personalized aesthetics via contrastive learning. First, a generic image aesthetics assessment network is pre-trained to learn the common aesthetic features. Then, image sets rated to have the same aesthetic levels by different users are employed to learn the differentiated aesthetic features through multiple level-wise contrast learning based on the generic aesthetic features. Finally, a target user's PIAA model is built by integrating generic and differentiated aesthetic features. Extensive experiments on four benchmark PIAA databases demonstrate that the proposed MTCL model outperforms the state-of-the-arts. |
author2 |
School of Computer Science and Engineering |
author_facet |
School of Computer Science and Engineering Yang, Zhichao Li, Leida Yang, Yuzhe Li, Yaqian Lin, Weisi |
format |
Article |
author |
Yang, Zhichao Li, Leida Yang, Yuzhe Li, Yaqian Lin, Weisi |
author_sort |
Yang, Zhichao |
title |
Multi-level transitional contrast learning for personalized image aesthetics assessment |
title_short |
Multi-level transitional contrast learning for personalized image aesthetics assessment |
title_full |
Multi-level transitional contrast learning for personalized image aesthetics assessment |
title_fullStr |
Multi-level transitional contrast learning for personalized image aesthetics assessment |
title_full_unstemmed |
Multi-level transitional contrast learning for personalized image aesthetics assessment |
title_sort |
multi-level transitional contrast learning for personalized image aesthetics assessment |
publishDate |
2023 |
url |
https://hdl.handle.net/10356/171829 |
_version_ |
1783955552363085824 |