Lei Feng news network, IBM has taken an important step in Watson visual identification image classifier (Watson Visual Recognition) processing ability was updated, it allows users to understand the image or video content. This system built-in tens of thousands of visual labels, so that the vocabulary is 2.5 times larger than the previous model. The expansion of the built-in vocabulary makes it greatly improve the ability to recognize special visual concepts.
Newly implanted visual tags cover a wide variety of visual concepts, including objects, people, places, activities, scenes, and other words about a particular feature, such as a particular color.
Of course, some companies have their own custom data, they want to create their own classifier for these data. Watson visual recognition also has the characteristics of custom development and classification. When the service needs to learn a new set of image tags for a field (like a product portfolio), developers can quickly develop a new custom model by providing sample images. The application can then use a custom model, combined with the most basic services, to describe the picture with a general description of words and domain specific descriptions. Custom classifiers can also be upgraded by adding new sample images.
Hill said the development of visual identity is an important step in IBM's ability to continuously improve Watson cognitive areas. It is based on the continuous development of visual understanding worldwide. The visual understanding made a series of breakthroughs, including the analysis of improved treatments for skin cancer patients using image, improve the image automatic captioning technology and artificial intelligence and creative breakthrough limit, making the world's first step of cognitive movie trailers etc..
If you want more information, you can clickHere, get more information about visual recognition services. Lei Feng network will also do follow-up concern for this research.