Researchers at Fujitsu and also the MIT Center for Brains, Minds, and Machines (CBMM) have achieved a “major milestone” within the quest to bolster the accuracy of AI models tasked with image recognition.

As described in an exceedingly new paper presented at NeurIPS 2021, the collaborators have developed a way of computation that mirrors the human brain to enable AI that may recognize information that doesn’t exist in its training data (also called out-of-distribution data, or ODD).

Although AI is already used for image recognition during a range of contexts (e.g. the analysis of medical x-rays), the performance of current models is extremely sensitive to the environment. the importance of AI capable of recognizing ODD is that accuracy is maintained in imperfect conditions – for instance, when the attitude or light level differs from the pictures on which the model was trained.

Improving AI accuracy

MIT and Fujitsu achieved this feat by dividing deep neural networks (DNNs) into modules, each of which is liable for recognizing a special attribute, like shape or color, which is analogous to the way the human brain processes visual information.

According to testing against the CLEVR-CoGenT benchmark, AI models using this system are the foremost accurate seen to this point when it involves image recognition?

“This achievement marks a significant milestone for the long-run development of AI technology that would deliver a replacement tool for training models which will respond flexibly to different situations and recognize even unknown data that differs considerably from the first training data with high accuracy, and that we anticipate to the exciting real-world opportunities its release,” said Dr. Seishi Okamoto, Fellow at Fujitsu.

Dr. Tomaso Poggio, a professor at MIT’s Department of Brain and Cognitive Sciences, says computation principles inspired by neuroscience even have the potential to beat issues like database bias.

“There could be a significant gap between DNNs and humans when evaluated in out-of-distribution conditions, which severely compromises AI applications, especially in terms of their safety and fairness. The results obtained to this point during this research program are a decent step [towards addressing these types of issues],” he said.

Going forward, Fujitsu and therefore the CBMM say they’ll try and further refine their findings in a trial to develop AI models capable of constructing flexible judgments, with a view to putting them to figure in fields like manufacturing and treatment.

Leave a Reply

Your email address will not be published. Required fields are marked *

You May Also Like

How to Transfer Nothing Black Dot NFT to Wallet & verify on Nothing Discord Server

The Nothing technology startup has announced the Nothing Community Dot NFT. The…

Nothing Phone (1) New Update Nothing OS 1.1.3 brings New Features, Camera Improvements and Bug Fixes

The Nothing Phone (1) is the most hyped smartphone of 2022. This…