CONNECT WITH US

Open AI Fab develops automated image matting models for advertising designers

Mark Tsai, Taipei; Adam Hwang, DIGITIMES Asia

Open AI Fab, a provider of subscription-based consulting in AI technology and AI solution, has developed Auto Hair, an AI-based automated image matting model to remove the background of hairs (human or animal) in images for use in advertising design, according to company CEO Jon Huang.

The development of Auto Hair was enabled by comparing open-source AI algorithms concerning image matting and integrating them as well as using trimap generated from masks of Google DeepLab v3+, Huang explained. The automatically generated trimap is an image consisting of foreground, background and an uncertain region, with pixels in the region classified as either foreground or background based on AI-based comparison for removing background. Auto Hair can reach fine removal of hairs in images, Huang said.

Open AI Fab has also used natural language processing (NLP) technology to build an AI-based model for law offices, Huang noted. The model can recognize food advertising with exaggerating texts, such as those about therapeutic effects, to avoid violation of regulations on advertising, Huang indicated. This is because lawyers' recognition is repeated and time-consuming work, Huang said, adding that the model can also indicate risky wording which is likely to violate regulations.

Open AI Fab plans to offer NLP-enabled chat robots in the second quarter of 2020, Huang noted. Such chat robots are expected to help users express moods and inner thoughts that are unable or inconvenient to be shared with others, Huang indicated.

Open

Open AI Fab CEO Jon Huang
Photo: Mark Tsai, Digitimes, November 2019