Panasonic Develops Image Recognition AI – Future Camera Innovation?

Artificial IntelligenceIndustry News

by Omri Keren LapidotJanuary 11^th, 2024

Panasonic Holdings Co. Ltd (Panasonic HD) has announced the development of a new AI algorithm system introducing a new classification algorithm that can handle multimodal distribution. As of writing this article, there are no real-world photographic implementations to speak of, but the tech itself is intriguing. Is a new Panasonic AI autofocus system coming?

AI, deep learning, and subject recognition

Subject recognition algorithms aren’t new. Their basic forms have been with us since before the dawn of mirrorless. Readers may remember Canon’s early attempts with DSLR-based iTR systems and how early camcorders and compact cameras could adjust settings based on shooting conditions, and there may be other instances where photographic devices used different sorts of recognition. The evolution continued and took an exponential leap forward with the mirrorless revolution. 

Mirrorless and AI – a perfect match

Mirrorless technology isn’t new. Digital cameras with some forms of live view have been around for over 20 years now. The significant development in recent years came with the exponential increase in data transfer and sensor readout speed. Cameras these days crunch through incredible amounts of data. AI and its more common variant – deep learning, thrive on big data. This convergence of a camera’s ability to provide a constant flow of live data and pre-learned algorithmic data enables significantly better interpretation of the scene. Subject recognition was just a matter of time, and we can now find it in most, if not all, current hybrid cameras, and even in some cine cameras.

What’s the big news then?

Well, at this level, we can’t be sure about real-world applications, but it seems that Panasonic is taking a relatively new approach: 

“The distribution of appearances within a category is actually not uniform. Within the same category, there are multiple subcategories with multiple different trends in appearance (multimodal distribution).”

(Panasonic Press release)

The multimodal distribution handling may potentially change the way subject recognition works. If you want to dive into the technical details, you’ll find them in the press release, but I’ll try to simplify them: The new system uses different AI algorithms that should recognize common subject characteristics even if their appearance is visually different. The same bird will look completely different when lit from the front or back, with clear or cloudy skies, not to mention inside foliage, etc. Two identical cars with different paint schemes look different even under similar lighting conditions. The new algorithms should be able to “look over” these variables and identify the subject based on several indications. 

How will Panasonic AI affect the photographic industry? 

In all honesty, it’s hard to determine. Too many variables are still unknown. Subject recognition is used in many different markets and applications, and we have no official confirmation as to whether or not it’s even coming to our cameras. Such a precise recognition system may elevate subject-detecting AF performance to a whole new level, but I find myself asking – How much better can autofocusing be?

Cameras these days easily achieve what was considered impossible just a few years ago. But there is one significant theoretical effect, and that’s the prospect of Panasonic leveling the playing field with Canon and Sony. Being a LUMIX user myself (S1R, S5, S5II) I’ve witnessed both the inferiority and the significant improvements Panasonic managed to implement, via firmware and hardware. If the company achieves comparable performance with its main rivals, the market will surely change. 

What are your thoughts concerning the advancement in AI recognition? Has the technical advancement in cameras changed your shooting technique for the better? Let’s discuss it in the comments below.