This is the reply I got from Digital Globe for the same query. I have inserted comments for other members in the community to follow through and hopefully this will open up the possibility for further discussion.
Thanks for reaching out with your questions. I'll try to answer below. Please follow up with any additional questions you have.
1) A number of the datasets have been open sourced on http://blog.tomnod.com. We are working on a better method for making data available. This will be through DigitalGlobe's open data website. We expect it to be completed soon. Once it is ready, some of the data collected in public Tomnod campaigns will be posted there.
2) We do provide a crowdrank score with each feature. This could be used to create a confidence matrix.
3) Agree that using this data to train automated algorithms is great use for the data. To do this you need access to both tomnod data and the imagery that tomnod features were extracted from. Often, getting access to the imagery is a challenge.
My comment: Training datasets can be of different resolution and there are existing methods to deal with imagery resolution
4) Contributions on the campaigns you see on tomnod.com are from volunteer users. We are clear in our terms that DigitalGlobe owns all contributions. We open source some, but not all of the data we collect. Here are a few reasons why we choose to or not to open source data:
My comment: Included in SAR agreement not in the Terms of Service Agreement
Partners that ask us to run tomnod campaigns often ask us not to open source the data. A good example is GlobalXplorer who is looking for the world's archeological sites. We do not open source the data because we do not want looters to use the data for destructive activities.
My comment: This is case specific
Often the challenge we face is more on open sourcing access to imagery. You can imagine much of the data we collect on disasters in tomnod (locations of flooded bridges for example) is only useful if you have access to the image that the Tomnod data was collected from. Open sourcing imagery (which is very expensive for DigitalGlobe to collect) is often a challenge.
My comment:Note open source datasets at 10m resolution and 5 m resolutions are available
All of this said, I am hopeful DigitalGlobe will start open sourcing more data soon. I will send you access to our open data site when it is released. This will be a much more organized and clear location for everyone to access the data that we do open source.
Thanks for reaching out and let me know if you have any more questions!