Will it be cheap to generate Splat outputs with this to use in Spark?
Atlas: A World Model for Spatial Intelligence
51–60 of 64 posts
Re: Atlas: A World Model for Spatial Intelligence
#52China, DeepAtlas please!
Re: Atlas: A World Model for Spatial Intelligence
#53Unfortunately I can see law enforcement and legal systems using this to make "crime scene recreation" videos that are believable enough to convince jurors despite being pure fabrication.
well, just to illustrate to the jury that the "recreation" is just an illusion built by prosecution, the defense should be able to present the same video with the lead prosecutor in the role of the perpetrator of the crime.
Re: Atlas: A World Model for Spatial Intelligence
#54Any plans for characters or is this strictly environments with the basic motion you see in the video?
Re: Atlas: A World Model for Spatial Intelligence
#55It is so good and probably only become better from here but for now I can see things appearing and disappearing (aka inconsistent hallucinations). See the dominos falling scene, it is easy to see dominos appearing and disappearing into thin air.
Re: Atlas: A World Model for Spatial Intelligence
#56Looks very very interesting. For my immediate use-case to simulate a real-world 3D space, would it work with exact dimensions? Like can I give it exact dimensions and then can it render it truthfully? For example, I want to say I have walls of 8ftx5ft, and I have N paintings of so-and-so specific sizes (I have the exact sizes), is it able to use the dimensions exactly?
Re: Atlas: A World Model for Spatial Intelligence
#57The blog post doesn't seem to mention what strikes me as the most interesting application of a model like this, namely extracting semantic information from its latent space. It mentions robotics applications, but only in the context of generating realistic world models for simulation. If you have a robot deployed in an environment, generating synthetic views of the environment you're in doesn't have any obvious value…
I am not sure why creating synthetic view of the environment should not have any value.
When you plan your path to go somewhere you can imagine how it would look like based on what you know and when new information comes in that contradicts it it gets updated to what it's really like.
The robot could do the same thing. It could create images/views of what path it wants to take and start acting on it without knowing the real situation yet.
Granted, the 3D representation that atlas also provides is probably more useful than just an image but the idea remains the same, we are giving robots a kind of idea how the current world around them most likely is based on previously learned data of other locations
Re: Atlas: A World Model for Spatial Intelligence
#58Assuming you had a 1:1 scale, what would be the difference between you taking a camera of an environment and measuring everything the same way vs. it being generated artificially in a 3d space?
Wouldn't the inputs be more or less the same wrt training a model with the synthetic data vs live video and the data captured there?
Re: Atlas: A World Model for Spatial Intelligence
#59Looking forward to the next step, where the reconstructed/hallucinated environment is dynamic and has a physics model etc
Re: Atlas: A World Model for Spatial Intelligence
#60This is incredible. One potential application that I'm thinking about already is the rapid iteration of video-game map blocking. Being able to drop in some 'initial state' configuration and then have it procedurally generate a handful of alternative configurations could make rapid prototyping a significantly quicker experience, especially if you wanted to see what a potential end result could look like. Furthermore,…
This is completely useless for video game map blockout. You need collisions for level blockout, which this does not provide. It makes gaussian splats, not geometry.