The authors present a novel approach to determine the optimal processing window for laser powder bed fusion (LPBF) additive manufacturing (AM) of metallic materials. The conventional method of printing 3D test samples and conducting metallurgical analysis is time-consuming and costly, especially for novel materials. In this work, the authors propose a solution using an unsupervised convolutional neural network to automatically generate the process map for the alloy based on single-track experiments. The proposed method utilizes unsupervised pretraining with the Invariant Information Clustering (IIC) loss \cite{Ji2018} and supervised fine-tuning to generate an accurate process map, reducing the time and cost associated with traditional methods. The track morphology is used as a form of process optimization to minimize defects and improve the overall quality of printed alloys and ultimately, metallic components fabricated using metamaterials. This is a proven method of reducing certain defects in the AM process \cite{Scime2019, Scime2019A} Thes results show that the melting mode of different alloys can be predicted by the trained network from just the top down view with reasonable accuracy, enabling the creation of useful process maps.