<p>Laser powder bed fusion for metals (PBF-LB/M) is an advanced additive manufacturing process in which metal powder is selectively melted and fused layer by layer. Maintaining consistent meltpool conditions is critical to ensure the production of fully dense, defect-free parts. While process parameters, such as laser power or scan speed, are usually held constant during a build, effective variations in meltpool conditions arising from geometry, thermal history, and energy redistribution during the build can lead to porosity formation defects such as lack of fusion or keyholing porosity, which are often detected only after fabrication. To address this challenge, in-situ monitoring techniques have been implemented to observe meltpool dynamics in real time. During fabrication process for nickel alloy 625 samples, high-speed infrared and off-axis imaging systems capture grayscale images of the process, which are converted into contour maps highlighting regions of interest, including the meltpool, meltpool vicinity or plume, and spatter. These images are analyzed using deep learning (DL) methods to detect potential anomalies. In particular, convolutional neural networks (CNNs) are employed for feature extraction and classification of meltpool conditions. The CNN model is trained on a labeled dataset of meltpool images and validated on a separate test set. It achieves a classification accuracy of 97.87% across three condition categories: acceptable normal meltpool, overheated meltpool and its vicinity, and underheated meltpool and its vicinity. This trained DL model is ready to be implemented in a real-time monitoring framework that enables effective detection of conditions resulting in defects during the build process and supports dynamic adjustment of process parameters, such as energy input, to reduce defect formation. By enabling real-time feedback and control, the system would enhance the meltpool consistency during the PBF-LB/M process.</p>