Justin900429 / computational-aesthetics

Python version for extracting computational aesthetics features

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

README

Introduction

This repo is to extract the computational aesthetic features introduced from the paper - "The Pictures we Like are our Image: Continuous Mapping of Favorite Pictures into Self-Assessed and Attributed Personality Traits". Some of the features had been removed and readapted. See the table below.

Feature Synopis

Category Name dimension
Color HSV statics
Emotion-based
Color diversity
Color name
5
3
1
11
Composition Edge pixels
Level of detial
Average region size
Low depth of field (DOF)
Rule of thirds Image size
1
1
1
3
2
1
Textural Properties Gray distribution entropy
Wavelet based textures
Tamura
GLCM - features
1
12
3
3

The description column is omitted. To see what each feature does, please refer to the paper. Additionally, there are some changes being made.

  1. The Faces features was removed.
  2. The realization of color diversity is different from the paper.
  3. The GIST descriptors was removed.
  4. The GLCM - features used only gray image and left only 4 features.

Installation

Install requirements

$ pip install git+https://github.com/Justin900429/computational-aesthetics

Import file

from CA import CA
...
# Create objects
img_path = "..."
ca = CA(img_path)
res = ca.compute_ca()
...
# update image path
new_path = "..."
ca.update(new_path)
new_res = ca.compute_ca()
...

See example for more details

Citation

@ARTICLE{
  7378902,
  author={Segalin, Crisitina and Perina, Alessandro and Cristani, Marco and Vinciarelli, Alessandro},
  journal={IEEE Transactions on Affective Computing},
  title={The Pictures We Like Are Our Image: Continuous Mapping of Favorite Pictures into Self-Assessed and Attributed Personality Traits},
  year={2017},
  volume={8},
  number={2},
  pages={268-285},
  doi={10.1109/TAFFC.2016.2516994}
}

About

Python version for extracting computational aesthetics features

License:MIT License


Languages

Language:Python 100.0%