Facebook publicly shares open source A.I & # 8211; Segmentation Object in image

Good news for programmers! Today Facebook has announced to the community a few open source intellectual software (AI) that help classify objects in images (Segmenting objects within images). Tools like The DeepMask, SharpMask , and MultiPathNet are now available on GitHub.

This is not the first time Facebook has offered these programs to the community. Currently, the entire research on artificial intelligence (Artificial Intelligence Research – FAIR) is open to the software making community can experience and build directly on it.

Image segmentation has "evolved" a step further, it can represent people, places, or objects in an image, and it can even confirm locations through data on images. To do that, Facebook is using an AI technology called deep learning, a technology that enables "training" intelligence through a lot of data networks and thereby enabling them to think from data sources. new material.

Briefly talking about these tools, DeepMask will create initial object masks, SharpMask will clarify these masks, and eventually MultiPathNet proceeds to identify the set objects from the masks, ”said the scientist of FAIR research. , Piotr Dollar has said on his blog.

This is not the first time that Facebook's AI system has been released. They had previously given Torchnet a pitch in June.

Other giants like Apple , Baidu , Google , and Microsoft , are also pouring a lot of money into researching and developing this deep learning technology.

The research labs of these companies vigorously compete in this segment, possibly to names like COCO . Good research will produce good applications, which can attract more users and gather more data.

Facebook also knows how to improve their applications with these tools. Dollar said:

Leaving the machine able to identify objects on the photos themselves, it will make it easier for users to search for specific images. This tool will also help visually impaired people to understand the content of photos their friends are sharing, the system will be able to help them do it easily.

Moreover, this is also a orientation to improve the user experience even more. Our next challenge is to apply this technology to video, where objects move continuously, interact, and change over time. We also have a number of specific directions with computer vision techniques that help interact on video in real time, understand and classify objects such as cats or food. And in particular, this classification technology will increase the interaction on Live videos.

ITZone via Venturebeat

Share the news now