Ensuring stability and designing suitable excavation and support patterns require precise rock evaluation. Discontinuities in rock are critically considered during rock assessments. This study proposes a method for acquiring three-dimensional trace mapping data using a Convolutional Neural Network and utilizing this data to characterize the orientation of rock discontinuities based on trace data. The three-dimensional trace mapping data is generated using simple image processing method without complex computations. To characterize orientation, Principal Component Analysis is used to calculate the normal vectors of the three-dimensional trace data, which are then used to determine dip and dip direction. The trace mapping results using CNN showed high trace persistence, with minimal omissions. The trace-based orientation characterization results were compared with surface-based characterization, confirming its effectiveness as an orientation characterization method.