/ImageAestheticAssessmentPyTorch

Image Aesthetic Assessment in PyTorch with implemented popular datasets and models (possibly providing the pretrained ones).

Primary LanguagePythonApache License 2.0Apache-2.0

Image Aesthetic Assessment in PyTorch

Image Aesthetic Assessment in PyTorch with implemented popular datasets and models (possibly providing the pretrained ones).

Note that this is a work in process.

Supported Datasets

  • AVA Datasets A large-scale database for aesthetic visual analysis
    • 250,000 images, 235556 (training) and 19926 (testing)
    • aesthetic scores for each image
    • semantic labels for over 60 categories
    • labels related to photographic style
  • AADB Datasets A aesthetics and attributes database
    • 10,000 images in total, 8458 (training) and 1,000 (testing)
    • aesthetic scores for each image
    • meaningful attributes assigned to each image (TODO)
  • CHAED A chinese handwriting aesthetic evaluation database
    • 1000 Chinese handwriting images
    • diverse aesthetic qualities for each image from three level
    • rated by 33 subjects

Supported Models

  • Unified Net Image Aesthetic Assessment Based on Pairwise Comparison – A Unified Approach to Score Regression, Binary Classification, and Personalization (In Progress)
  • PAC-Net PAC-NET: PAIRWISE AESTHETIC COMPARISON NETWORK FOR IMAGE AESTHETIC ASSESSMENT (TODO)
  • NIMA NIMA: Neural Image Assessment

How to use

python -u train.py --config path/to/your/config