The behaviors of an individual are usually regulated by subjective valuations that are not only driven by external information but also influenced by internal affective state. The default mode network (DMN) has been demonstrated to play a crucial role in monitoring affect, but for subjective valuation, the level of processing at which the DMN is situated remains unclear. In this study, we developed an artificial neural network (ANN) that predicts participants’ valuation of oil paintings, and examined the relationship between ANN representations and brain activity. We found that the visual areas V1 and V2 are highly correlated with shallow layers of the ANN, while the default mode network is correlated with deeper layers. Vision-to-value processing was well-aligned with the principal gradient of cortical connectivity. These findings suggest that the default mode network plays a central role in processing subjective value, and that subjective value is imputed by projecting external information onto an internal space via the principal gradient.