Rethinking Image Color Aesthetics Assessment: Models, Datasets and Benchmarks

Shuai He, Yi Xiao, Anlong Ming, Huadong Ma

Beijing University of Posts and Telecommunications

Introduction

Our refined work of ICCV 2023 work DeT, weights and dataset can be downloaded from:

• DeT_plus_ICAA,
• DeT_plus_PARA,
• DeT_plus_SPAQ,
• ICAA17K+ dataset,
• RN50.

Details will be published after the acceptance of the paper, we aspire for our work to make a valuable contribution to the ongoing research on ICAA within the community!

ICAA17K+

To enhance the ICAA17K+ dataset, we have incorporated 2,000 detailed labels concerning color attributes including colorfulness, harmony, and temperature annotations.

Performance of DeT-Plus

We develop a comprehensive benchmark comprising of 17 methods, which is the most extensive to date, based on three datasets (ICAA17K+, SPAQ, and PARA) for evaluating the holistic and sub-attribute performance of ICAA methods. Our work achieves state-of-the-art (SOTA) performance on all benchmarks.

Requirement

einops==0.6.1
ftfy==6.1.1
nni==2.10.1
numpy==1.25.2
pandas==2.1.0
Pillow==10.0.0
PyYAML==6.0.1
regex==2023.8.8
Requests==2.31.0
scikit_learn==1.3.0
scipy==1.11.2
setuptools==65.5.1
tensorboardX==2.6.2.2
timm==0.9.7
tqdm==4.66.1
yacs==0.1.8