LightBlog

lundi 22 janvier 2018

You Decide Honor’s Next AI Camera Feature

Lately Honor has been very active in the community. They have stayed long-past the end of scheduled events to respond directly to your questions in the forums with experts from their team.  In the previous live Q&A, you asked AI related questions that were addressed by an Honor Product Manager; some of those questions were taken to CES and answered by a Microsoft AI expert.

Honor continues to interact with the community to learn and innovate in ways no other company in doing. Soon we will be holding another event with Honor, where you will help decide what AI powered scene is added to Honor’s intelligent scene recognition. Stay tuned for more details about this upcoming event. In the meantime, check out this video, which does a great job of explaining the Honor View 10’s intelligent camera.

Here are some answers from the last Q&A:

Q: Where do you see the future of photography going with the help of Honor’s AI assisted camera technology?

A: The NPU is the key to make the AI computing possible in honor View 10. With the NPU, honor View 10 provides 13 different scene recognition when taking pictures. The phone knows what type of scene you are shooting and will smartly adjust the best photographic parameters for your shooting to make sure you can easily take the best pictures.

Q: How dose NPU capability differ from a CPU or other AI enabled CPU in the market?

A: Smartphones are different from servers in many respects, including size, power supply, power consumption, and heat management – all of which place restrictions on smartphone chipset design. Mobile SoCs have to deliver the best possible performance, while ensuring the highest possible density of core functions with optimal use of energy. The Kirin 970’s development team designed this chip on a solid foundation of past breakthroughs, integrating the world’s first NPU (Neural Network Processing Unit) specifically designed for a mobile device. This specialized processing unit, combined with Huawei’s innovative HiAI mobile computing architecture, delivers an AI performance that far surpasses any CPU/GPU build. Compared to a quad-core Cortex-A73 CPU cluster, the Kirin 970’s new heterogeneous computing infrastructure delivers 25x the performance with 50x greater efficiency. Simply put, the Kirin 970 chipset – with its embedded NPU – can perform the same AI computing tasks faster and with far less power. In a benchmark image recognition test, the Kirin 970 processed 33 images per second, which was several times greater than other chips on the market.

Q: Does NPU help in RAM and resource allocation during heavy usage?

A: The NPU is more focused on speeding the AI tasks, just like the GPU is more focused on graphic processing, while the RAM and resource allocation are mainly handled by the CPU. It is for the NPU that the Kirin 970 in honor View 10 keeps providing a stable AI ability even when the system load is high. A normal smartphone without an NPU will have to face a drop on AI processing in heavy loaded situations.

Q: What are the capabilities of the View 10’s AI processor? Will it learn our behavior, getting smarter and smarter? How does it change the way we use phone? How will this AI technology revolutionize mobile usage?

A: 1)The Huawei Kirin 970 chipset brings you a faster, stronger, and seamless smartphone experience. Its built-in NPU (neural-network processing unit) enables deep learning based on user behavior, ensuring your Honor View 10 truly understands you.

2)Because of AI technology, EMUI 8.0 understands how you use your Honor View 10 and allocates resources by anticipating your behavior. It’s always working in the background to make sure your Honor View 10 runs smoothly even after months of use.

The Honor View 10 features the Kirin 960 NPU, bringing powerful AI to the camera.

The NPU in the Honor View 10 enables the camera to use AI to determine different types of scenes adjust the settings accordingly.



from xda-developers http://ift.tt/2DoPbzX
via IFTTT

Aucun commentaire:

Enregistrer un commentaire