Intro

If we come across an animal we assume to be a “cat” while walking down the street, how do we know it is a cat and not a tiger? We subconsciously collect all available visual information, including the body size, ear shape, whether it looks wild or not, the length of its tail, and the size of its claws, and comprehensively decide it is not a tiger—to our relief. Will a computer embedded with AI recognize a cat in the same way as a human? With the emergence of the era of big data and the rapid increase of photo, video and image data, the amount of interest in and research on “computer vision,” a technology on “machine perspectives,” has been explosive. Development in deep learning technology, in particular, has driven rapid advances in research on computer vision, leading to a large number of significant research results. AI is growing ever more sophisticated, going beyond simply “seeing” something to “understanding” and responding to it. Virtual human technology has integrated various technologies, including face generation and speech synthesis, and has become a new way for AI to interact with clients in a variety of fields, including finance, public services, and education, as it is recognized as “quasi-sentient” in that it not only delivers images but can also offer real-time communication.

What is
AI Vision?

There are six main areas of research on computer vision: Detection, which identifies objects in images; Segmentation, which involves dividing an image into different sets of pixels; Recognition, which identifies the object being sought; Classification, which involves separating objects according to type; Enhancement, which improves image quality; and Restoration, which recovers lost images. One of the most prominent areas where computer vision technology, armed with big data and artificial intelligence, can have a dramatic impact is in the financial sector. This sector, which handles and processes huge amounts of data in a variety of fields, is the ideal area to maximize automation and efficiency by using computer vision technology for document and image data processing. Major examples include automatic recognition, classification, and analysis of imaged documents, customer information, and financial products.

Virtual human technology is a key representative of AI Vision capabilities and enables natural interaction with the user in a variety of media environments that support digital financial services. This technology is emerging as a next-gen financial communication tool to provide the user with customized explanations of complex financial processes through dialogue, and offering intuitive access and understanding to users needing additional help, such as the elderly and non-Koreans. Recent trends to adopt this technology can be seen not only in the financial sector but also across other industries, including education, healthcare, and public services. It is becoming an advanced  user interface, further innovating the customer experience.

Future Research

HIT researches and develops a variety of AI Vision technologies, including document detection and automatic extraction, and virtual human video generation. A major area of research is the development of a core content extraction service using Optical Character Recognition (OCR) and Natural Language Processing (NLP) for recognition and handling of documents. There are countless potential applications of OCR technology in the financial sector, as it can maximize convenience and effectiveness as business process automation lowers related costs in time and money. We are also working to secure the original technology that can provide more intimate remote services by converging Real-time Talking Face Generation and AI-based conversation technologies. HIT is committed to technological research and development of services in all areas that enable financial service innovation by utilizing AI Vision technology, and aims to position itself as a leader in AI Vision technology, not only within the Hana Financial Group, but also in the global financial sector.