Data for the paper "Learning to generalize to new compositions in image understanding", Yuval Atzmon, Jonathan Berant, Vahid Kezami, Amir Globerson and Gal Chechik

Project page

The format of the *split* text files is Triplet ID|Image ID|S, R, O

Triplet ID can be ignored.

License

Dataset was derived from annotation of the COCO dataset, and therefore we release it under the same license, which is Creative Commons Attribution 4.0 License.