3D Software

Physna launches Thangs – a deep learning-based 3D model search engine

Geometric deep learning specialist Physna has launched what it is calling the world’s most powerful geometric search engine – Thangs.

Instead of scanning for text or images, Thangs uses deep learning algorithms to index 3D models based on the polygons, or triangles, that make up their volumes. At launch, the free-to-join site already had more than a million public 3D models indexed, with plans to expand in the coming months.

Beyond just an intelligent search algorithm, Thangs also provides version control functionality and ‘compatible part predictions’ for the 3D community, and is expected to be a hit with designers, engineers, and 3D enthusiasts.

Paul Powers, CEO of Physna, explains: “Thangs is Physna’s first open product, but it uses some of the same powerful technology at the core of our enterprise offerings. Our enterprise product is widely recognized as the technological leader in 3D search and analysis, and we’re democratizing some of its capabilities through Thangs. This product is built to be powerful enough for a leading aerospace CAD engineer, yet simple enough for literally anyone to use.

Users are able to search for and comment on 3D models in the Thangs ecosystem. Image via Physna.
Users are able to search for and comment on 3D models in the Thangs ecosystem. Image via Physna.

The Google x GitHub of 3D models

Physna is calling its product the 3D world’s Google x GitHub crossover and looking at the project’s leadership team, it’s not hard to see why. Dennis DeMeyere, Physna’s CTO, was a former technical director at the Google Cloud CTO’s office. In July, Jason Warner, GitHub’s current CTO, also joined the company as a board member. With the all-star line-up, Physna is aiming to incorporate major functionality features from these two technology giants into its own 3D model platform.

Thangs was built on the premise that the 3D data world is still a highly manual one when compared to traditional software development. There is no central search engine so finding the files you need is a tedious process. Collaborative work within companies is also rudimentary due to a lack of version control software, and may involve cloud storage or sharing via email.

How does it work?

With Thangs, users are able to upload parts and receive suggestions for where that part may be used and what commercially available components may be compatible with it. There is, of course, a traditional text-based search box to find desired parts as well. Users can search based on the object’s physical properties, measurements, and features and receive predictions about its function, cost, materials, and performance.

The site also functions in a social capacity, as designers and colleagues are able to share and collaborate on 3D models seamlessly. Version control is automated – much like GitHub – and users are able to leave comments and ‘like’ models to save them for later, which triggers a notification for the uploader. Since a 3D designer’s portfolio of work is accessible from their profile, it can also serve as a resume of sorts.

According to Powers, Thangs is on track to become the world’s largest 3D model database within a year of its launch. It indexes any public models it finds automatically and as it is based on a deep learning algorithm, it gets more and more sophisticated with every addition.

The search result for the word 'engine'. Image by 3D Printing Industry.
The search result for the word ‘engine’. Image by 3D Printing Industry.

Machine learning in additive manufacturing is becoming increasingly more common as 3D printing technology advances. Earlier this month, engineering firm Renishaw partnered up with robotics specialist Additive Automations to advance automated post-processing technology for metal 3D printed parts. The collaboration will involve using deep learning algorithms to automatically detect and remove support structures using robotic arms.

Elsewhere, researchers from Argonne National Laboratory and Texas A&M University have developed an innovative new machine learning-based approach to defect detection in 3D printed parts. With the help of real-time temperature data, the scientists were able to make correlative links between thermal history and the formation of subsurface defects during laser PBF. Now, the team plans to develop the work with more data sets and an improved machine learning model.

The 4th annual 3D Printing Industry Awards are coming up in November 2020 and we need a trophy. To be in with a chance of winning a brand new Craftbot Flow IDEX XL 3D printer, enter the MyMiniFactory trophy design competition here. We’re happy to accept submissions until the 30th of September 2020.

Subscribe to the 3D Printing Industry newsletter for the latest news in additive manufacturing. You can also stay connected by following us on Twitter and liking us on Facebook.

Looking for a career in additive manufacturing? Visit 3D Printing Jobs for a selection of roles in the industry.

Featured image shows Physna’s Thangs UI. Image via Physna.