Google AI can create 3D images from 'flat' 2D images.

DeepMind, a UK-based company recently developed a Generative Query Network (GQN) - a neural network designed to teach artificial intelligence (AI) how to imagine things that will look like How from another side.

Specifically, AI will analyze 2D images and then render 3D views. It is worth mentioning that the artificial intelligence of DeepMind only looks at at least 3 images, not using any input data or knowledge store at all, then can immediately predict the 3D version of the image What will happen.

AI researchers are working hard to teach machines how to perceive as human beings, making assumptions after observing their surroundings, but specifically guessing the shape of objects and landscapes in particular. but it has not been "admired".

For example, you take a photo of a rubic block and ask Ai to recreate the image from another angle. Artificial intelligence - using GQN, must figure out what the rubic block (light, shadow, straight lines of the block) will look like on the other side to produce the image we need. .

Currently, DeepMind's artificial intelligence system has not been trained with the image of the real world. Next, the researchers will render it to render the real life landscape from the image.

Deepmind's GQN system is expected to use only 2D images that can produce extremely accurate 3D scenes in the future.

See more:

1. Summary of online AI courses for free
2. 6 steps to start learning artificial intelligence programming (AI)
3. Review important milestones in the history of more than 60 years of artificial intelligence development
4. AI can identify the gender of men and women through smiles