Recent advances in full-field measurement techniques have enabled convenient acquisition of full-field displacement data under multiaxial deformation. Such rich deformation data facilitate non-parametric identification of material behavior without relying on a prescribed constitutive equation or an assumed mathematical material model, thereby alleviating the need to formulate new constitutive models for emerging materials. In this context, a recent study demonstrated that physics-informed neural networks (PINNs) can predict nonuniform mechanical fields in a mesh-free manner and identify constitutive behavior for homogeneous materials, even from incomplete full-field measurements with missing points. PINNs incorporate governing equations and boundary conditions as soft constraints in the loss function, enabling data-driven learning while enforcing physical consistency. In engineering practice, heterogeneity commonly arises in composites and layered/coated systems, and assuming homogeneity can lead to non-negligible errors in mechanical-field prediction and material identification. However, simultaneously inferring the material layout and material response for multimaterial structures remains a largely open and challenging non-parametric problem. Therefore, in this work, we propose a PINN-based framework that simultaneously identifies the material layout and material response of multimaterial structures from displacement data obtained via numerical virtual experiments. The proposed framework consists of (i) a neural network that approximates the displacement field, (ii) a module that estimates the deformation gradient, (iii) a network that predicts the spatial distribution of constituent materials, and (iv) networks that learn material behavior. In particular, we employ material-wise separated networks in the constitutive learning stage to independently train distinct material responses. Numerical examples demonstrate that the proposed method robustly predicts both material layout and material behavior not only from complete full-field displacement data but also under incomplete observations with missing points.