Skip to content

CV: Image classification with a few plot twists.

Notifications You must be signed in to change notification settings

dx2-66/art-challenge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 

Repository files navigation

ML Art challenge

This challenge is aimed on creation of AI that capable to determine a genre of an artwork based on its visual appearance. Visual appearance of an artwork is represented by photos that are carefully distributed into categories.

This was a wonderful opportunity to try and implement error correction code approach to classification. Rather than creating a number of separate binary classifiers, the neural network was designed to be trained on class codes rather than plain classes.

An elegant non-iterative self calibrating solution to probability estimates for such a classifier from Reducing multiclass to binary by coupling probability estimates by Bianca Zadrozny was used.

This solution scored 4th place on a public test and 12th place on a private test at https://codenrock.com/contests/masters-of-arts#/

Units required:

  • Pandas
  • PyTorch
  • skorch
  • timm
  • scikit-learn
  • PIL
  • NumPy
  • Matplotlib
  • Seaborn

About

CV: Image classification with a few plot twists.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published