Skip to content

yemrekara/crowd_labeling

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 

Repository files navigation

Crowd Labeling For Continuous Valued Annotations

The code in this repository is a reimplementation of the work described in "Modeling annotator behaviors for crowd labeling" [1] and "Actively estimating crowd annotation consensus" [2]. The original code for the former work was implemented in MATLAB and the latter, a more comprehensive one, was implemented in C++. However, the C++ code is part of a larger project that I am unable to share.

I reimplemented the main part of the code in Python 3.7.3 for reaching a wider range of researchers from various disciplines. The code is unevitably slower than the original C++ implementation. It currently runs on a single core; parallelizing the parameter and consensus update codes should decrease the inference time. Note that the precision of the results may slightly vary due to implementation and hardware differences.

You are encouraged to play around with the code and apply it to your own data for continuous-valued consensus estimation in crowd labeling.

If you use this work, please do not forget to cite the related papers given below. In addition to the papers below [1,2], the detailed descriptions of the models and their derivations can be found in my PhD thesis [3].

– Yunus Emre Kara (yekara@yekara.com)

http://yekara.com

References

  1. Kara, Y. E.; Genc, G.; Aran, O.; & Akarun, L. (2015). "Modeling annotator behaviors for crowd labeling", Neurocomputing, 160, 141-156.
  2. Kara, Y. E.; Genc, G.; Aran, O.; & Akarun, L. (2018). "Actively Estimating Crowd Annotation Consensus", Journal of Artificial Intelligence Research, 61, 363-405.
  3. Kara, Y. E. (2018). "Crowd Labeling For Continuous Valued Annotations", PhD Thesis, Bogazici University, Istanbul, Turkey.

Datasets

About

Crowd Labeling For Continuous Valued Annotations

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published