With the recent increase in the number of social robots, expectations from robot technology are increasing daily. In particular, the development of cognition technology using deep learning has enabled various services, leading to the expansion of the robot market. However, in a dynamic real environment, robot object recognition requires the concept of intelligence rather than performance. In this study, we constructed an object recognition intelligence evaluation model based on the visual perception of humans. Then, we evaluated the object recognition intelligence of a social robot using a scenario based on this model.