Portuguese 
English 
Spanish 
3 min of reading 
0 comments 
Scientists Created a Compact Camera with Innovative Optics, Accelerating Object Identification with Greater Speed and Energy Efficiency
Scientists from two of the leading universities in the United States, University of Washington and Princeton University, developed a revolutionary compact panoramic camera for computer vision.
Using an innovative approach, the new prototype reduces energy consumption and promises to identify objects at an impressive speed: the speed of light.
Computer Vision with a New Camera
Computer vision is a field of artificial intelligence that enables computers to reflect on and interpret images and videos.
-
A mass of polar air is crossing Southern Brazil towards the Southeast and will cause something meteorologists call a thermal shock; the temperature could drop 10 degrees in just a few hours, and anyone who isn’t prepared will be caught completely by surprise.
-
BYD dethrones Volkswagen and becomes the best-selling brand in Brazilian retail just four years after setting foot in the country. Traditional automakers are in despair and are asking the government for urgent barriers to curb the Chinese invasion that threatens to destroy R$ 103 billion of the national auto parts chain.
-
Cargo planes full of machines will depart from China heading to Ceará to supply the construction of TikTok’s megadata center, a project of more than R$ 200 billion that aims to become the largest in Brazil and generate thousands of jobs.
-
The company that wants to resurrect the mammoth now targets an antelope extinct 260 years ago and needs more than 100 genetic alterations to bring back an animal that only exists in five museum specimens.
Traditionally, computer vision systems rely on electronic processing done by conventional hardware. However, researchers have found a way to perform part of this processing directly in the optics of the camera.
Arka Majumdar, a professor of electrical engineering and physics at the University of Washington, explains the innovation: “This is a completely new way of thinking about optics, very different from the traditional. The optics are designed along with the computational block, allowing transmission to be carried out optically.”
Instead of using a conventional glass or plastic lens, the new camera utilizes layers of 50 metalenses, which are lightweight, flat lenses composed of microscopic nanostructures.
These structures manipulate light and function as an optical neural network, an AI system inspired by the workings of the human brain.
Practical Advantages and Energy Efficiency
One of the main advantages of this approach is speed. The camera operates with processing done at the speed of light, allowing for identifying and classifying images more than 200 times faster than traditional neural networks, which use conventional hardware.
And it’s not just the speed that impresses; accuracy is comparable to the most advanced neural networks. Additionally, energy consumption is significantly reduced, as the camera’s optics use the incoming light to function, unlike traditional systems that rely on electricity.
Felix Heide, a professor at Princeton University, highlights the broad potential of this research. He mentions that applications can range from autonomous vehicles to medical devices and smartphones.
“Today, every iPhone has AI or vision technology. This work is still in its early stages, but it could benefit these devices in the future.”
The Impact on the Development of Autonomous Vehicles
The study, published in the journal Science Advances, presents a significant advance in optical computing.
With this integration of technology with the camera’s optics, the researchers developed a computer vision system capable of performing neural network calculations during image capture, even before recording them on the camera sensor. This enables real-time analysis with impressive accuracy.
The scientists are also looking to the future, with the expectation that the technology will play a crucial role in the navigation of autonomous vehicles.
As part of the next stage of research, they plan to test the prototype on more complex datasets and on problems that unlock greater computational power, such as object detection.
This type of technology is essential for autonomous cars to identify obstacles and make real-time decisions.
-
1 person reacted to this.