Download and resources

Training dataset

The training dataset is made of 15000 scenes of 100 frames each, approx. 25 hours of videos (split in 4 parts, 36GB each):

Participants are free to augment this training with any other training data (provided it is explained in the submission paragraph)

Dev dataset

The development dataset is made of 30 scenes per block (4.1GB in one archive):

Test dataset

The test dataset,is made of 1080 scenes per block (one archive per block, about 49GB each):

Note

The block 4 is not released yet.

Starting kit

All you need for the challenge participation is here:

Additional resources

  • BibTex citation of the reference paper for the dataset:

    @article{DBLP:journals/corr/abs-1803-07616,
      author    = {Ronan Riochet and
                   Mario Ynocente Castro and
                   Mathieu Bernard and
                   Adam Lerer and
                   Rob Fergus and
                   V{\'{e}}ronique Izard and
                   Emmanuel Dupoux},
      title     = {IntPhys: {A} Framework and Benchmark for Visual Intuitive Physics
                   Reasoning},
      journal   = {CoRR},
      volume    = {abs/1803.07616},
      year      = {2018},
      url       = {http://arxiv.org/abs/1803.07616},
      archivePrefix = {arXiv},
      eprint    = {1803.07616},
      timestamp = {Mon, 13 Aug 2018 16:47:10 +0200},
      biburl    = {https://dblp.org/rec/bib/journals/corr/abs-1803-07616},
      bibsource = {dblp computer science bibliography, https://dblp.org}
    }