Open Source

31692 readers

388 users here now

All about open source! Feel free to ask questions, and share news, and interesting stuff!

Useful Links

Rules

Posts must be relevant to the open source ideology
No NSFW content
No hate speech, bigotry, etc

Related Communities

Community icon from opensource.org, but we are not affiliated with them.

founded 5 years ago

MODERATORS

[email protected]

Infinigen (infinigen.org)

submitted 4 days ago by [email protected] to c/[email protected]

6 comments fedilink hide all child comments

Infinigen is a procedural generator of 3D scenes, developed by Princeton Vision & Learning Lab. Infinigen is optimized for computer vision research and generates diverse high-quality 3D training data. Infinigen is based on Blender and is free and open-source (BSD 3-Clause License). Infinigen is being actively developed to expand its capabilities and coverage

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 12 points 4 days ago (1 children)

That’s cool. I’d love to see that turned into a game, just to explore the scenes.

[–] [email protected] 2 points 3 days ago* (last edited 3 days ago) (1 children)

I would like it to be possible to give some hints to the model by providing maps showing where roads, buildings, vegetation should be and let it work its magic. Maybe take it further by procedurally generating buildings based on their footprints and building type, like hospitals , schools etc, as identified in maps. That way we can have racing sims in the roads we are familiar with. And much more...

[–] [email protected] 1 points 3 days ago (1 children)

As this isn't based on AI, interpretation of existing scenes could be...hard?

But you could try to build a process chain, where AI sets the parameters for infinigen, after analysing an existing model, and then it should probably be able to fine tune and compare or something

But I'm not to deep in AI development to really speak here

[–] [email protected] 2 points 3 days ago

There exist maps derived from satellite images called Land Use Land Cover maps that categorise each pixel into predefined classes like built up, forest, water, roads etc. Granted that these days semantic segmentation is used to generate such maps but traditional image processing and digitization has always been used traditionally. There is no AI involved in using them for the purpose I mentioned though. Gaming engines like unity have built in tools as well as add-ons like Gaia that have cool procedural generation features. Using such maps in conjunction could help in creating realistic and familiar worlds and this blender tool gives me much hope.