Nvidia has recently released a new case study examining the performance of its artificial intelligence-based learning techniques, known as Percent-Scale PnF (Proportional-Size Neural Filtering).
This technology, which may have applications in self-driving cars, autonomous robots, and cybersecurity, uses an innovative approach to quickly process vast amounts of data and learn from it. In particular, this technology drastically reduces the amount of data needed to create a useful model.
Nvidia’s research analysts focused specifically on the ability of its AI technology to quickly learn features and identify objects in images. To test this capability, they created a custom image-recognition dataset consisting of nearly 6,000 labeled, full color images. This dataset was used to train the AI to recognize the differences between objects, such as a car, a person, and a tree.
The AI system was then given new, untrained images to identify and classify them into the correct categories. To test its performance, the researchers compared the accuracy of the machine’s results with the accuracy achieved by using a traditional convolutional neural network (CNN).
Using Percent-Scale PnF, the Nvidia AI system achieved a classification accuracy of 91 percent. This result exceeds the accuracy of the CNN-based system, which achieved 76 percent. Furthermore, the new system uses far less processing power, requires less memory, and can be trained much more quickly, making it an ideal solution for applications with time and resource constraints.
The success of Percent-Scale PnF technology demonstrates Nvidia’s commitment to creating smarter, more efficient AI systems. This new technology provides a powerful solution for recognizing and classifying objects in images, with many more potential applications in the future.