Panasonic has joined the world of AI. And it seems that it hasn’t only jumped on the bandwagon but has led the charge with a new development. Their latest creation is an image recognition AI that sheds the limitations of its predecessors. It can tell apart the objects that appear differently depending on their surroundings.

Panasonic recently published a paper on its research, the result of a collaboration between Junpei Goto, Yohei Nakata, Kiyofumi Abe, and Yasunori Ishii of the Panasonic HD Technology Division, and Takayoshi Yamashita, a professor at Chubu University. They performed the study under the guidance of experts as part of the Panasonic Group’s AI expert training program called REAL-AI. They also published a paper on the new technology, explaining it more in-depth.

The main point of Panasonic’s new model is to recognize and understand the nuances of different images. For example, it can tell apart “a Labrador retriever basking in the sun” from “a shivering Chihuahua caught in the rain.” A dog is not simply a dog – it matters what breed it is and in what context it’s put. It tackles the real-world challenge of multimodal distribution, where objects within the same category have diverse appearances due to different factors. These can include lighting, orientation, and background.

Seeing through the noise

Training the algorithm, generally, often encounters a diversity issue. You show it pictures of robins perched on branches, but what happens when it encounters a hummingbird hovering over a flower? Or a penguin waddling on ice? Traditional AI would be lost, unable to reconcile these vastly different appearances with the concept of a “bird.” This is where Panasonic steps in.

Panasonic’s AI doesn’t just memorize textbook definitions. Instead, it creates a two-dimensional map of all the possible ways an object can appear. This map, like a detailed atlas of bird variations, allows the AI to handle unexpected situations like soaring eagles or tiny chicks in their nests.

The applications

You may wonder: Where can we use this? Panasonic notes that this technology can ultimately lead to safer roads. It would allow autonomous vehicles to navigate complex environments more accurately, identifying pedestrians and traffic signs under any conditions. Robots in factories could also identify and sort objects even if they’re damaged or obscured, which would result in boosting efficiency.

Panasonic isn’t the first major company to join the AI craze. Adobe launched its own image generator, Firefly, in early 2023. They also added a bunch of generative AI features to programs like Photoshop, Lightroom, Premiere Pro, and After Effects. Then there’s Meta, Samsung (whose AI creates everything), and Google, with Bard. Of course, we can’t forget the biggest players: Midjourney, StabilityAI, and OpenAI, which are primarily devoted to artificial intelligence.

It’s also important to remember that this is still an emerging technology with technical hurdles to overcome. However, Panasonic’s research seems like a major leap forward, paving the way for a (terrifying) future where AI truly sees the world in all its complexity.

[via Photo Rumors]