Macnica President (headquartered in Yokohama, Kanagawa, Japan; : ; capital: 11,194 million yen; hereinafter ), a total service and solution provider in semiconductor, network, cyber security, and AI/IoT, and ALBERT Co. (ALBERT, headquartered in Shinjuku-ku, Tokyo; Takeshi Matsumoto, ; capital: 1,022 million yen) have announced the launch of a new library utilizing multimodal AI to automate the generation of image captions. President Kazumasa Hara Macnica
This will enable users to easily use the library, which combines multifaceted information and is similar to human judgment, thereby expanding the introduction of multimodal AI in users' business domains and accelerating the social implementation of AI.

Macnica 's AI expert organization, AI Research & Innovation Hub (ARIH), released the Re:Alize^*1 service in September 2020 as part of the service portfolio of "macnica.ai", which enables technology verification in the shortest possible time. We have licensed and provided libraries ^*2 that are essential for AI implementation so that they can be implemented and operated effectively. Prior to this, Macnica and ALBERT entered into a capital and business alliance agreement in November 2019, and both companies have been promoting initiatives to solve social issues using digital data.
The multimodal AI we are providing is the first library jointly developed by Macnica and ALBERT.

As the knowledge and understanding of AI deepens, and more companies are working to create new value through implementation, improving the accuracy of AI has become an issue. (Fig. 1)

 
図１（出典：AI白書2020 p.308 図3-4-23「AIを利用する上での課題」）

“Multimodal AI” is an AI that can process and judge compositely, such as image recognition and voice recognition, which are classified as “specialized AI” widely used in current business fields. This is expected to bring us closer to the problem-solving ability of an expert who integrates and judges a variety of information.
This time, among the multimodal AI, we will develop and start providing a library that can learn image and text data and automatically generate image captions. By providing it as a library, customers can perform technical verification and start operation in the shortest time with a clear understanding of the period and cost.