This work is a natural progression from a lot of other prior work in the literature... but that doesn't make the results any less impressive. The examples shown are amazingly, unbelievably good! Really GREAT WORK.
Based on a quick skim of the paper, here is my oversimplified description of how this works:
During training, an agent navigates an artificial 3D scene, observing multiple 2D snapshots of the scene, each snapshot from a different vantage point. The agent passes these snapshots to a deep net composed of two main parts: a representation-learning net and a scene-generation net. The representation-learning net takes as input the agent's observations and produces a scene representation (i.e., a lower-dimensional embedding which encodes information about the underlying scene). The scene-generation network then predicts the scene from three inputs: (1) an arbitrary query viewpoint, (2) the scene representation, and (3) stochastic latent variables. The two networks are trained jointly, end-to-end, to maximize the likelihood of generating the ground-truth image that would be observed from the query viewpoint. See Figure 1 on Page 15 of the Open Access version of the paper. Obviously I'm playing loose with language and leaving out numerous important details, but this is essentially how training works, as I understand it based on a first skim.
EDIT: I replaced "somewhat obvious" with "natural," which better conveys what I actually meant to write the first time around.
I, literally just 15 minutes ago, had a chat with a friend of mine exactly about how what we are doing right now with computer vision is all based on a flawed premise (supervised 2D training set). The human brain works in 3D space (or 3D+time) and then projects all this knowledge in a 2D image.
Here I was, thinking I finally had thought of a nice PhD project and then Deepmind comes along and gets the scoop! Haha.
I don’t think this is a novel idea, but it is still a great topic for a PhD. While the results in this paper look impressive, my suspicion is that the system doesn’t generalize particularly well. (I suspect this from experience with similar, albeit simpler, ideas, as well as from looking at the datasets.) If you can make a system that generalizes to new environments and objects, or a system that works with real-world natural image/video data, that would be a tremendous accomplishment.
Generalization is a more fundamental problem, and (imho) should be tackled first at a more fundamental level.
For example, if you have a classifier that can recognize cats, it doesn't mean it will work for cartoon cats. You'd have to train the system all over again with cartoon cats. Instead, you want the system to learn more like humans, where only a small number of examples is necessary to make the connection between real and cartoon cats.
It is possible that the problems are related—-it may be that, to achieve human-like generalization, neural nets need to learn in a human-like environment, instead of from a folder full of images. But time will tell.
This has been said many times in different ways over the years. To achieve human-like intelligence one needs a human-like body operating in a human-like environment. It's the first of the E's in: embodied emergent extended enactive. https://plato.stanford.edu/entries/embodied-cognition/
Maybe in trivial examples, but bleeding edge has for quite some time adopted 3D convolutions (either 3D space or 2D + time), and combined them with RNNs for more power.
The trick is to then to have one agent try to take on the perspective of another agent, then change its behavior accordingly (depending on task goal). I.e. theory of mind.
It may be splitting hairs, but I think the mammalian brain, at least, can simulate/remember/imagine additional 'dimensions' like X/Y/Z spin, derivatives of velocity like acceleration/jerk/jounce.
Is space 11 dimensional (M string theory) or 2 dimensional (holographic principle)? What 'dimensions' does the human brain process? Is this capacity innate or learned; should we expect pilots and astronauts to have learned to more intuitively cognitively simulate gravity with their minds?
> This work is a somewhat obvious progression from a lot of other prior work in the literature... but that doesn't make the results any less impressive.
Everything takes figuratively forever to train, and the field is moving incredibly fast, so everything you see is both directly adjacent to previous work and also impressive.
To me, this represents the best of science--that we can collectively make rapid progress without having to invoke an Einstein figure to make some magical leap!
Based on a quick skim of the paper, here is my oversimplified description of how this works:
During training, an agent navigates an artificial 3D scene, observing multiple 2D snapshots of the scene, each snapshot from a different vantage point. The agent passes these snapshots to a deep net composed of two main parts: a representation-learning net and a scene-generation net. The representation-learning net takes as input the agent's observations and produces a scene representation (i.e., a lower-dimensional embedding which encodes information about the underlying scene). The scene-generation network then predicts the scene from three inputs: (1) an arbitrary query viewpoint, (2) the scene representation, and (3) stochastic latent variables. The two networks are trained jointly, end-to-end, to maximize the likelihood of generating the ground-truth image that would be observed from the query viewpoint. See Figure 1 on Page 15 of the Open Access version of the paper. Obviously I'm playing loose with language and leaving out numerous important details, but this is essentially how training works, as I understand it based on a first skim.
EDIT: I replaced "somewhat obvious" with "natural," which better conveys what I actually meant to write the first time around.