That is a good suggestion. We are trying to set up a system where people can register to use the images. We were at the CGIAR meeting a few weeks ago where there was much talk of open and big data. One line of arguments asked should data like the images we invest in and curate be made available without subsequent curation. The only reason our images are useful is that they were expertly curated- first because we collected them in a research center where multiple experts work and secondly that we as experts arranged them. But the danger is that other groups using ML might not keep the metadata with the images or have a mis-labelling event ultimately introducing bias. The images would be available but we would know for whom and what purpose. That way we could ultimately query the resultant classification tool to see if it has a bias.

We are still evolving and changing so if people have views, share them here