Accurate crop growth prediction is a key requirement for sustainable vertical farming systems, with particular focus on Adaptive Vertical Farms (AVFs). These next-generation vertical farms rely on an adaptive configuration proposed in [1], where vertical shelf spacing is dynamically adjusted according to crop growth over time, as sketched in Figure 1. This contribution builds upon our previous results on crop growth prediction for adaptive vertical farms [2] and extends them by investigating two complementary modeling paradigms. Crop growth is assessed through dry weight, a reliable measure of biomass that is less affected by transient environmental changes. The first paradigm is based on a dynamic model, whose parameters are identified from measurements to provide interpretable predictions. The second paradigm relies on a data-driven approach based on artificial neural networks trained directly on experimental data. The two approaches are compared in terms of prediction accuracy, robustness with limited data, and suitability for integration within optimization and control strategies for adaptive shelf scheduling. Numerical experiments are conducted using both simulated scenarios and real measurements collected from vertical farming campaigns. The obtained results provide practical guidelines for selecting modeling tools in sustainable agricultural technology (widely known as "AgriTech") applications. [1] Ghio M., System for the vegetable garden and nursery cultivation of plants, Italian Patent, no. IT20200011161, 2020. [2] Chnib E., and Bagnerini P., Gaggero M., Zemouche A., Parameter Estimation of a Dynamic Growth Model for Lettuce in an Adaptive Vertical Farm, Proc. IEEE 20th International Conference on Automation Science and Engineering, pp.1169-1174, 2024.