In this series, journalists, Chika Nishida, report on the latest trends in the digital industry.The theme this time is the security camera "Google Nest CAM" released by Google in Japan.We approach how Google is enhancing this product.
Click here for the last link
The "security camera" from the past was literally a "camera".It was a human being to decide what was reflected in the video that was visible from the location where the camera was installed.
However, it is not limited to the fact that Amazon and Google are provided by Amazon and Google for smart homes, and Apple is installed in the "Camera linked to Homekit".
It is image recognition technology that holds the key.AI determines what is in the image, records only the necessary parts, and is notified to smartphones and the like if they need to be warned quickly.
It doesn't make sense to keep recording like a security camera in a store, and it is desirable to notify you of your change as needed.Therefore, a "consistent mechanism" that recognizes what is in the camera in the "on -device AI" built into the device and notifies smartphones via the Internet service.
Such a part is a platform business that major IT companies, such as Amazon, Google, and Apple, are good at.Every company has a "paid internet service" for security enhancement and privacy enhancement.In addition to selling hardware, it is a market that is expected to sell online services for a fee.
In that case, the differentiated points of each company are "how to enhance the service" and "How to increase the accuracy of AI".
In the case of Japan, Google is the only major IT company that has entered this genre, but the technology polished in the US market is trying to overwhelm the home security camera market in Japan.
The interesting thing is that you use "other than reality" to grow AI.Google is also affiliated with Unity, which develops graphic engines for games.Unity creates a large amount of scenes that can be considered inside and outside the house as a realistic CG, and the accuracy is improved by using it as information for AI to learn.
The amount of CG used for learning is more than 25 million copies, and in order to prepare a large amount of images in a short time, it was necessary to use CG as well as real images.is.It is also important that if you use CG, you do not need to collect images while paying attention to infringing someone's privacy.
You can clearly see that the current CG can realize the realism that is close to reality.The idea of utilizing such elements that seem to be different in fields seems to show Google's flexibility.
Weekly GETNAVI, here is the back number
続きを読む