Quantifying the grain size and orientation is essential for understanding the properties of the materials, especially when increasing the creep resistance of high-temperature alloys such as Haynes 282. The grain distribution of additively manufactured metal is attributed to various factors, such as processing parameters and part geometry, making it difficult to predict the microstructure. Haynes 282 being an FCC single-phase structure, polarized light methods cannot be applied to segment out grains accurately. Therefore, it is challenging for other imaging methods to segment and analyze the orientation of the grains. In this work, we present a cost-and-time-efficient pipeline for predicting grain size and orientation distributions. This method uses a convolutional neural network (CNN) that segments out grains in SEM images and predicts a relative grain orientation, which can work as an alternative for time-consuming EBSD.