This seems like a great way of publishing public datasets. However, as someone who works on a computer vision startup I don't think I could really use this. I

This seems like a great way of publishing public datasets. However, as someone who works on a computer vision startup I don't think I could really use this. In my work data annotation, visualization, and versioning cannot be easily separated. The effort we would need to put in to use quilt might be better spent building a simple versioning system on top of our current data infrastructure.