Common Objects in Context
Source
I think that the COCO data comes from this site: cocodataset.org/#download
Description of the Data
The homepage for the COCO data set states "COCO is a large-scale object detection, segmentation, and captioning dataset."
We can use this dataset to demonstrate parallelization techniques. It is just the right size to be nearly impossible to do without parallelization, and also to get the "wow" factor when students get it working!