Digital Cameras, yes also the ones embedded in smart phones, have advanced functions for image recognition. The objective of those functions is to set the ideal focus and exposure to get a good photo. An experienced photographer might do a better job (although in cases where decision time is very short, like capturing a rapidly moving object, the automatic systems works better than your fingers….) but in general the average user will find that the “intelligence” in the camera is pretty good and worth exploiting.
These embedded functions have grown over time in capabilities, now being able to detect our face and the eye closest to the lens (that is important because that is the point where you want the camera to focus) and more recently are able to detect a few animals faces as well. Some can even detect a smile and shoot to capture it. Others can wait to take the shot till eyelids are open… Amazing, isn’t it?
Camera makers label their cameras as “smart”, “intelligent”. Some flatly state that they are using AI and in a way they might be right because they have been using AI algorithms to learn from hundreds of thousands of photos taken by professional photographers what is the best exposure and have acquired the capability to compare a situation in real time with one of those “learnt” exposures.
Now Sony has announced a new image sensors on a chip, the IMX500, that embeds AI algorithms that can learn as they are used. An application could be developing, at sensor’s level, the capability to recognise specific objects in a warehouse, or a store, to facilitate inventory. Sony plans to release this chip along with SaaS, Software as a Service, this latter offered at an additional price. For what I was able to understand from the announcement there might be two ways to grow the intelligence in the sensor: one based on the capability of the sensor itself (with its AI support circuits and software) to learn, the other to use Cloud based SaaS to train the AI algorithms. Sony announced a partnership with MS Azure to deliver Cloud based services and this can be seen now as a building block of this strategy. It is not clear if the end customer (a device manufacturer that will be using and training the chip to fit a specific application) will have to pay both Sony and MS to exploit the chip capability or just Sony.
All in all, this chip may become a game changer in some AI applications given its embedded capability to support AI.