To calculate the wall condensation under a wide range of flow types for efficient engineering calculations, this study proposes a wall condensation model based on the neural network. A series of representative COPAIN condensation cases were selected to construct a training dataset based on the resolved boundary layer (RBL) method. These cases encompass various flow types, ranging from low-speed natural convection to high-speed forced flows, as well as different working pressures, steam mass fractions and condensation temperatures. A wall condensation model is then trained using well-validated CFD data obtained from simulations on a very fine mesh. Finally, the developed neural network model is deployed in the CFD code, and its performance is evaluated by comparing it with experimental results, RBL results, and existing wall function model predictions across a broader test set of COPAIN condensation cases. The results demonstrate that the neural network model exhibits excellent generalization capability, accommodating diverse flow types. The novel neural network-based wall condensation model provides a new approach for calculating wall condensation with coarse boundary grids and offering an innovative methodology for constructing wall function models.