| --- |
| license: other |
| task_categories: |
| - image-classification |
| - image-regression |
| task_ids: |
| - facial-beauty-prediction |
| language: |
| - en |
| tags: |
| - beauty-prediction |
| - face-analysis |
| - facial-beauty |
| pretty_name: SCUT-FBP5500 |
| size_categories: |
| - 1K<n<10K |
| --- |
| |
| # SCUT-FBP5500 Dataset |
|
|
| ## Dataset Description |
|
|
| The SCUT-FBP5500 dataset is a diverse benchmark for facial beauty perception test. It includes 5500 frontal faces with diverse properties (male/female, Asian/Caucasian, ages) which are rated with beauty scores ranging from [1, 5] by 60 volunteers. |
|
|
| ### Dataset Summary |
|
|
| - **Number of instances**: 5500 facial images |
| - **Demographic breakdown**: |
| - 2000 Asian females |
| - 2000 Asian males |
| - 750 Caucasian females |
| - 750 Caucasian males |
| - **Beauty scores**: Range from 1 to 5 (average of 60 raters) |
| - **Facial landmarks**: 86 landmarks per face |
|
|
| ## Dataset Structure |
|
|
| - `image`: The facial image |
| - `beauty_score`: Beauty score (average rating from 60 volunteers) |
| - `race`: Race of the person (Asian or Caucasian) |
| - `gender`: Gender of the person (Male or Female) |
| - `image_name`: Original image filename |
| - `has_landmarks`: Whether facial landmarks are available for this image |
|
|
| ## Dataset Splits |
|
|
| The dataset provides: |
| - Standard 60/40 train/test split |
| - 5-fold cross-validation splits (available in the original dataset) |
|
|
| ## Citation |
|
|
| ``` |
| @article{liang2017SCUT, |
| title = {SCUT-FBP5500: A Diverse Benchmark Dataset for Multi-Paradigm Facial Beauty Prediction}, |
| author = {Liang, Lingyu and Lin, Luojun and Jin, Lianwen and Xie, Duorui and Li, Mengru}, |
| jurnal = {ICPR}, |
| year = {2018} |
| } |
| ``` |
|
|
| ## License |
|
|
| The dataset was collected for research purposes. Please contact the original authors for commercial use. |
|
|
| ## Source |
|
|
| This dataset was created by South China University of Technology. |
|
|
| For any questions about this database, please contact the authors by sending email to lianwen.jin@gmail.com and lianglysky@gmail.com. |
|
|
