https://github.com/ivan-zhou/personalized-photo-recommendation
This repository includes work for the RecSys conference paper Aesthetic Features for Personalized Photo Recommendation
https://github.com/ivan-zhou/personalized-photo-recommendation
Last synced: 2 days ago
JSON representation
This repository includes work for the RecSys conference paper Aesthetic Features for Personalized Photo Recommendation
- Host: GitHub
- URL: https://github.com/ivan-zhou/personalized-photo-recommendation
- Owner: Ivan-Zhou
- Created: 2018-08-23T06:38:36.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2018-10-04T18:16:34.000Z (about 7 years ago)
- Last Synced: 2025-06-08T11:50:22.998Z (5 months ago)
- Language: Python
- Homepage: https://arxiv.org/abs/1809.00060
- Size: 23.4 KB
- Stars: 2
- Watchers: 0
- Forks: 2
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
## Personalized Photo Recommendation
This repository includes work for the **Aesthetic Features for Personalized Photo Recommendation**.
### Data
You need to save your data in the folder `data/`. If you only have one dataset, you can save under the folder with name
'sample/'. If you have multiple datasets, you can save multiple them in separate folders with any name format.
Under each folder (for one datset), you have a folder named 'validation/' for parameter tuning, 'test/' for
evaluation, and `photos` to save all photos to be used. Under each of the folder, there will be `matrix_train.npz` for training and `matrix_test.npz` for testing.
The photo data should be all in `.jpg` format.
### Create Aesthetic Feature Embedding
#### Create Color Embedding
You can create color embedding for all your photos with the script below:
```
cd aesthetic_features
python create_color_histogram.py --data-folder data/
```
By default, the program will create RGB, HSV, and HLS embedding respectively for all photos.
#### Create Style Embedding
You can run style embedding for all your photos with the script below:
```
cd aesthetic_features
python create_style_embedding.py --data-folder data/
```
### Evaluate Models
You can evaluate with all the models on your data with the script below:
```
python model_evaluation.py --data-folder data/ --task test
```