Diffgram Launches Open Source Training Data Platform for Machine Learning


The Diffgram Training Data Platform is now Open Source – after three years of closed source work. Diffgram is similar to closed source offerings like Labelbox but is different in that it’s Open Source.

The Open Source Diffgram community is already over 600 technical people strong. There are people submitting code for new cloud providers, creating bug reports for issues, and more. Overall, there is a global team creating a shared repository of knowledge. That means companies can rest assured they will always be able to get support and updates.

This software is important to all companies, especially large companies, therefore it’s important to show leadership in this space. “We estimate, based on analytics from installs and public information, that Diffgram has now been installed on more systems than the nearest competitor, Labelbox. Considering that each installation of Diffgram can service hundreds, and sometimes even thousands of users, this is a substantial presence,” reads a company statement.

There are many software packages in this general space that are open source – however Diffgram team said they are the first to put the entire platform in the open. The platform that’s supported over 40,000,000 annotations on Diffgram.com is now Open Source.

“There is no separate remote service to get value from it. Instead, Diffgram’s business model is to charge larger teams who have a significant volume of users and people who wish to use the fully hosted platform.”


Please enter your comment!
Please enter your name here