Machine learning: azure seing ai should recognize and label pictures of pracation

Machine Learning: Azure Seing Ai should recognize and label pictures of pracation

Microsoft’s Azure Ai Cognitive Services have presented a new algorithm for Image Captioning, which is to approach human ability to accurate image description according to the first test. According to blog entry, Microsoft plans to offer its automatic image description KI system as a standalone software outside of Microsoft products, in the Azure Cloud. The camera app "Seeing ai" Provide the first Microsoft product an update, the implementation of the algorithm in products of Office Suite such as PowerPoint, Outlook and Word should follow according to the blog entry of the developer team.

Camera app describes visible objects in text form

The app "Seeing ai" Microsoft published in 2017. As a assistant app, she uses the camera’s eye of the smartphone to describe the world’s blind and surprised the world. According to the publisher, it can identify and associate household items, read texts and scan, describe scenes and even recover friends. But also pictures from the virtual world, other apps, e-mails and social networks can be grasped and described with Seing AI. So there is movement in the field of image recognition with Ki, especially in the event of images with alternative text.

For people with a frosted visibility, the good news could be: about that "old"-Attribute receive pictures a description in text form to allow people with visual loss (or a weak internet connection) the perception of visual content on the internet, in documents and social media. The alternative text can be output for example with readers. In the realitat are the old attributes, which were allowed to allow all accessible access to visual content, often empty or incorrectly faulty.

User view of assistant apps like Seing Ai

The AppleVis community, a group of blind and visualized internet users existed since 2012, has 2019 SEEING AI "Best app" for people with impregnated visuality. Since publication of the App 2017, this expert community recommended the app several times as a practical assistance tool (at least for users of Apple-advised with iOS operating system).

In the Blog Entry of the Microsoft Developer Team, it is called for the updated version with the new algorithm that this is twice as accurate as the previous image labeling system that Microsoft has been developing since 2015. In advance, the developers of the Microsoft Cognitive Services team had published test results in a scientific essay, which shows that the algorithm developed by the Azure Ai Cognitive Services on comparison values "human" Benchmarks approaches or the comparison brands in the test even slightly surpassed. However, such test results are added with caution because it is not about the actual image perception and sugariness by humans, but calculated comparative values, and the benchmarks of Nocaps probably cover only a part of possible visual concepts of human perception.

Potential of KI software for barrier freedom

Microsoft is not the first and not the only provider of apps and algorithms for image recognition and automated image caption. Thus Facebook 2018 introduced an (more) object recognition, which runs as open source project DetectRon. The underlying software is written in Python and adds a broader portfolio of the provider in the field of image analysis.

Google also works within the framework of its ml kit and the project Firebase at automated software for image recognition and possibly also image lettering. Last, Google’s Android ML Kit had been made by transcribing handwritten notes on the Digital Ink Recognition API, which should work in around 300 languages. Shortly before, Google’s ML developer had the Android ML Kit from its close connection to the ML Kit SDK from Firebase to concentrate with the new kit on KI applications for the mobile area. After the development jump from the ML department of Microsoft is exciting, with which Facebook and Google probably attract as a next.

More information about the new algorithm of Microsoft can be found in the blog entry of Cognitive Services. The feature is available on the Azure Cognitive Services. Interested parties can be informed about the AI pages of Microsoft on the App Seing AI, which is available in the App Store for download.

Like this post? Please share to your friends:
Leave a Reply

;-) :| :x :twisted: :smile: :shock: :sad: :roll: :razz: :oops: :o :mrgreen: :lol: :idea: :grin: :evil: :cry: :cool: :arrow: :???: :?: :!: