Context The [LabelMe][1] project has been run out of MIT for many years, and allows users to upload and annotate images. Since the labels are crowdsourced, they can be of poor quality. I have been proofreading these labels for several months, correcting spelling mistakes and coalescing similar labels into a single label when possible. I have also rejected many labels that did not seem to make sense. Content The images in the LabelMe project as well as the raw metadata were downloaded from MIT servers. All data is in the public domain. Images within LabelMe may have been taken as far back as the early 2000s, and run up to the present day. I have worked through 5% of the LabelMe dataset thus far. I decided to create a dataset pertaining to meals (labels such as plate, glass, napkins, fork, etc.) since there were a fair number of those in the 5% I have curated thus far. Most of the images in this dataset are of table settings. This dataset contains: 596 unique images 2734 labeled shapes outlining objects in these images 1782 labeled image grids, with a single number representing which portion of a grid cell is filled with a labeled object Acknowledgements Many thanks to the [people of the LabelMe project!][2] Inspiration I want to see how valuable my curation efforts have been for the LabelMe dataset. I would like to see others build object recognition models using this dataset. [1]: http://labelme.csail.mit.edu/Release3.0/ [2]: http://labelme2.csail.mit.edu/Release3.0/browserTools/php/credits.php