The overarching goal of this research work is to fabricate biocompatible, porous bone scaffolds that are not only mechanically robust but also dimensionally accurate for the treatment of osseous fractures, defects, and musculoskeletal diseases. In pursuit of this goal, the objective of the work is to develop an image-based intelligent platform, based on convolutional neural network, for prediction of the functional properties (such as porosity, stiffness, and compressive strength) of composite bone scaffolds (composed of polyamide, polyolefin, and cellulose fibers) fabricated using fused deposition modeling (FDM) process. FDM is a material extrusion additive manufacturing process, which has been extensively utilized for the fabrication of a wide range of biological tissues and constructs for tissue engineering applications. As a high-resolution method, FDM allows for deposition of composite materials with complex formulations as well as complex porous microstructures. Despite the advantages and engendered applications, the FDM process is inherently complex; the complexity of the process is, to a great extent, the result of complex physical phenomena (such as non-Newtonian material deposition, layer fusion, and phase change) in addition to unavoidable material-process interactions (e.g., molten polymer flow deposition and subsequent layer fusion vs. translation speed). Besides, there is a wide spectrum of scaffold design, composite material, and fabrication process parameters (such as molten polymer viscosity, scaffold morphology, nozzle diameter, deposition temperature, and forced convection rate influencing solidification rate) contributing to the complexity of the FDM process. As a result, investigation of the impact of consequential design, material, and process parameters as well as their interactions would be required for optimal fabrication of mechanically strong, dimensionally accurate, and porous composite bone scaffolds. In this study, an image-based convolutional neural network (CNN) platform is presented with the aim to intelligently learn the complex dynamics of composite material deposition and ultimately predict scaffold porosity. In this study, the CNN model is trained on the basis of monochromatic images acquired from FDM-fabricated bone scaffolds via a high-resolution charge-coupled device (CCD) camera. The bone scaffolds were fabricated based on a medical-grade composite material, deposited using a converging microcapillary nozzle having a diameter of 800 μm with a deposition temperature, translation speed, and layer height of 225 °C, 15 mm/s, and 400 μm, respectively. The CNN model is utilized for in-process prediction of the morphological properties of the fabricated bone scaffolds. Overall, the outcomes of this study pave the way for smart, patient-specific fabrication of robust and porous bone scaffolds with tunable medical and functional properties.