- A Generative Query Network (GQN) which can imagine objects from different angles, like humans.
- The AI enabled system can turn a 2D image into a 3D image.
- The GQN can also generate new aspects of the 2D image.
Lenses, dual lenses, HDR lenses, have all been trying to enhance the interaction of users with an image and things have just taken an exciting turn.
The researchers at DeepMind have come up with an AI-enabled system which can promisingly convert a 2D image to a 3D image. The system is called as a Generative Query Network (GQN) which can imagine objects from different angles just like humans.
The system is far more complicated than it sounds and the researchers have spoken about it in detail in their research paper. The GQN is programmed to predict the environment of the 2D image to come up with a 3D interpretation of the same. Not only is the system capable of thinking like humans, but it can also render the 3D objects without any prior tip on the various angles to imagine.
So, let us figure out.
How it Works :
Generative Query Network or GQN is an algorithm which can help to render 3D models of objects and scenes from 2D images which are called as Scene representation—the process of converting visual sensory data into concise descriptions and usually to do so, it requires large sets of labelled data. Scene representation is an essential attribute of intelligent behaviour.
The AI system has two parts:
- Representation network: Here, the images are converted into a code that the computer can understand. Thus, the image is now a code which the computer can understand precisely.
- Generation network: This network uses the code obtained from the above network to create the various other angles of the object which, in the initial images are unseen.
The GQN can learn and adapt to combine all the information to form an accurate 3D model. The network also possesses the ability to create scenes by itself without any prior input of data.
So Let’s Summarise :
Even if the system is in its budding stage, the scope is enormous! The algorithm is another new member of the continually growing family of AI enabled algorithms. Apart from merely converting images to 3D, the system can be used in multiple applications which as of now remain unexplored.