For activities in virtual space, which is in the spotlight as a new lifestyle of mankind, companies and research institutes around the world focus on creating objects inside the virtual space and creating and setting up animal and plant environments. Data objects in virtual space need to improve more realistic modeling and feedback and subsequent movements to maximize human five senses. Require two-dimensional specific object recognition as a preprocessing step for research to transfer real physical objects to virtual digital data. This study intended to improve the performance of prior work in the field of conversation, which needs to improve object recognition, by using a deep learning-based real-time image detection system. It was confirmed that the object recognition rate in the artwork was improved by targeting traditional Korean art that lacked relatively implementation data due to differences from the parameters of modern art. Research to increase accuracy in forming three-dimensional modeling should be conducted later.