The home service robot becomes a new dominant challenge for robotic researches especially for the development of the smart home system. The home service robot are used for nursing care service purposes which also designed to handle multiple types of house tasks such as garment folding. This work considers the problem of recognizing the configuration of the garment that already crudely spread-out on a flat platform before folding it into desired configuration. Later, the deformable model with its parameters is generated from the real garment for the dexterous robotic manipulation purposes. The approach to generate a parametrized 2D/3D deformable model for each category of real garments is proposed. Three garment categories which are towels, shirts and trousers are used as garment testing samples. The proposed approach will attempt to best fit the 2D mesh deformable model to the crudely spread-out garment in the captured image. The parameters of garment such as landmarks and robotic grasping points are predicted and the folding procedure can be planned. Once the parameters are defined, the 3D garment model such as shirt and trouser can be generated to provide a completed configuration for the real garment, where the mesh topology of the 2D garment model is preserved. Different mesh sizes can be considered by using this proposed approach. The experimental results shown that the approach can achieve effective and accurate on a set of the garment samples.