Google launches video intelligence API to automatically identify objects

According to foreign media TechCrunch, Google announced today that it will launch a new machine learning API at the Cloud Next conference in San Francisco. The API supports automatic recognition of objects in the video, making them searchable.

The API, called Video Intelligence, will allow developers to develop applications that automatically recognize objects in the video. So far, most similar image recognition APIs provided in the cloud focus on object recognition in still images, and with the help of this new API, developers will be able to develop applications that support searching and discovering information in video. program. For example, that means you can search for "puppy" or "flowers" in the video.

In addition to extracting metadata, the API helps mark scene changes in the video.

Those videos must be stored in Google's cloud storage service. Google has provided a demo for this, and developers can apply for a beta version.

In the keynote speech today, Google Cloud AI and machine learning chief scientist Fei-Fei Li pointed out that the world of pixels has extended beyond the image. Video has always been a challenge for machine learning researchers. But now the new service makes it as easy to extract information from the video as it is to extract information from the image.

In addition, the company's Cloud Machine Learning Engine is now open to the outside world. With this tool, developers can use the TensorFlow framework to create custom machine learning models.

As Li Feifei pointed out in today's keynote speech, Google wants to popularize the machine learning technology developed internally. The video intelligence API is another example.