Skip to content

A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning

License

Notifications You must be signed in to change notification settings

yordanh/clevr-dataset-gen

 
 

Repository files navigation

Modified CLEVR Dataset Generation for bi-object spatial relations

This is the code used to generate the data for the simulated experiments as described in the paper:

Disentangled Representations for Explaining and Learning from Demonstrations
Yordan Hristov, Daniel Angelov, Michael Burke, Alex Lascarides, Subramanian Ramamoorthy
Presented at CoRL 2019

Code for the baselines used in the paper can be found here.

If you find this code useful in your research then please cite

@inproceedings{hristov2019disentangled,
 title={Disentangled Relational Representations for Explaining and Learning from Demonstration},
 author={\textbf{Yordan Hristov} and Daniel Angelov and Michael Burke and Alex Lascarides and Ramamoorthy, Subramanian},
 booktitle={Conference on Robot Learning (CoRL)},
 year={2019}
}

The repository is a fork of the original CLEVR dataset. All modification were developed and tested on Ubuntu 16.04.

About

A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%