Learning how you, our users, interact with Fusion 360 has always been a big part of what we do, be it at the Fusion 360 Academy, listening sessions, our online forums, or through the interactions of our support team. The feedback we get from the community allows us to continuously improve Fusion 360, month after month, year after year. This direct input is critical, but we also have an opportunity to leverage more sophisticated technologies, like machine learning, to better understand your designs and usage patterns on a broad scale. For example, if we see repetitive steps being taken in the design timeline, we know there is an opportunity to help our users design faster and more efficiently.

We believe machine learning shows great promise to help learn how you design today, to improve our products of tomorrow. Machine learning is a technology that powers capabilities like autocomplete of text when typing emails and automatic labeling of photos to make them searchable. How might we help you autocomplete standard CAD components? Or help search for designs such as fasteners, enclosures, or brackets without having to manually tag each one? The technology and research to enable these scenarios are getting closer and closer, and we have already begun to bring a machine learning-based visual similarity search to Fusion 360. However, much depends on having good data to learn from.

Maintaining the integrity, confidentiality, and privacy of user data is second to none at Autodesk. We also believe we can establish responsible ways to work with user design data to improve your experience with Fusion 360. As a member of our research team, I’m excited to give you a glimpse of a project we’re working on that aims to do just that!

Today we are sharing with the community our plans to re-release designs shared publicly on the Autodesk Online Gallery website in a machine learning ready format. We hope the Fusion 360 Gallery Dataset, as we are calling it, will enable the broader research community to dramatically improve how machine learning can be applied to CAD. The designs included in the dataset will be only those that are already available for public download from the Autodesk Online Gallery. The native Fusion 360 designs will be converted into solid models, meshes, point clouds, and design metadata, such as the modeling operations used, for ease of use with machine learning. Rest assured that this re-release does not include design data stored in your Fusion 360 personal or Teams account. Additionally, the dataset will only include design data and not other personal information or comments shared via the online gallery.

We are excited to explore the future of CAD with our users and the research community. In the coming months, we will share our progress and highlight how the research community is advancing CAD through machine learning.

Try out Fusion 360 today by downloading your free trial below.