OpenAI Announces Point-E System to Create 3D Models Using AI
Every day there are more and more moments in our daily lives where Artificial Intelligence (AI) helps us, often without us even realizing it. This type of software, which simulates human behavior and even thinking, finds solutions to certain problems based on machine learning techniques and Deep Learning. Indeed, you may have seen them in personal online shopping, credit card fraud prevention or even as voice assistants. And it’s no wonder that AI is becoming more and more common, considering the results of a 2020 study conducted by Statista that found that the global artificial intelligence software market is expected to reach $126 billion by 2025. In fact , AI can also be used for producing 3D models! Or at least that’s what OpenAI is aiming for with its newly introduced Point-E system.
It is already known that companies such as the technology giant Google, for example, offer a service with its tool DreamFusion in which text can be converted into a 3D model. However, since this generation would take hours, it is probably all the more exciting that the company OpenAI, in which the American entrepreneur Elon Musk is significantly involved, presented their machine learning system called Point-E to the market this week. According to the company, this work can now be reduced from hours to minutes.
How does Point-E work?
It is possible to create 3D objects with the help of Point-E thanks to the generation of a large number of point clouds in a space, which more or less represent the 3D shape. As already noted, this system is supposed to work faster than other offerings on the market. This is also reflected by the ‘E’ in the name, which stands for efficiency.
Although it is possible to work faster with Point-E, there are also disadvantages. These point clouds can be easier to produce, they do not capture the fine shape or texture of the object. The manufacturer of the system was aware that this would be a significant drawback for many users, so an additional AI system was added to convert this point cloud into mesh. Mesh is a kind of catch basin of vertices, edges or even surfaces used for the final definition of an object. That said, even with mesh, some parts of the model can be overlooked, meaning that the shape cannot be represented in the truest sense.
Additionally, as already mentioned, Point-E also offers an image-to-3D model in addition to the text-to-image model. The first is a system trained to understand associations between words and their corresponding images. In the case of the image-to-3D model, on the other hand, images are generated in combination with 3D objects, enabling the system to obtain a more efficient understanding of them. For example, if you are a user of Point-E and you have a text request, such as that of a gear wheel, which must have a certain diameter and width, then the text-to-image function generates this 3D model , which starts with a rendered object, which is then transferred to the image-to-3D and generated as a point cloud.
You can imagine that this type of processing requires a large database of models, so Point-E is capable of performing these commands. The company OpenAI says that this technology for generating 3D models is not yet perfectly mature, but they strongly believe that this activity works faster than compared to previously existing techniques.
The benefits that OpenAI offers to its Point-E users
Of course, one or the other would immediately think of a whole range of possible applications of this system with AI. However, the team around OpenAI announces that their system of point clouds can be particularly well applied for the production of real objects. For this they call 3D printing. However, they say they are convinced that Point-E can also find its footing in the gaming and animation sector in the long term. If you want to learn more about OpenAI, please click HERE.
What do you think of OpenAI and its Point-E system? Let us know in a comment below or on our Linkedin, Facebook, and Twitter pages! Don’t forget to sign up here for our free weekly newsletter, the latest 3D printing news straight to your inbox! You can also find all our videos on our YouTube channel.
*Cover photo credits: OpenAI