Annotated IMB Levels

This dataset contains 1,437 IMB levels that were evaluated according to the following criteria: enjoyment, visual aesthetics, and difficulty. That is, users evaluated whether the levels are enjoyable to play, if they have good visual aesthetics, and also how difficult it was to complete them. The evaluation score of each level is the average score of different users on a 7-likert scale for each criterion. The score of 7 for enjoyment, visual aesthetics, and difficulty for a level means that the level is very enjoyable to play, it has excellent visual aesthetics, and it is very difficult to complete, respectively.

Click on the link below to download a copy of Marcus Persson's IMB (implemented in Java). You will find our dataset in a folder named labeledlevels within the IMB project. Run the class RunCustomized to test one of the levels in our dataset. Modify the class LevelSceneTest accordingly to test another level from our dataset.

Please cite the following paper if you use this dataset.

    author = {W. M. P. Reis and L. H. S. Lelis and Y. Gal},
    title = {Human Computation for Procedural Content Generation in Platform Games},
    booktitle = {Conference of Computational Intelligence and Games},
    pages = {99--106},
    year = {2015},

Download Dataset