Over the past two decades, facial modeling has received significant improvement. However this improvement, still face some limitations such as illumination, poses, variations of facial structure, low resolution and the last but not the least is that the variations are not linear for which Active Appearance Models (AAM) faces the challenges too. The research takes a look at the Deep Appearance Models(DAMs). DAMs are divided into three main steps. Two separate prior models are created for shapes and size. These models interactions are interpreted and finally, a fitting algorithm is used to synthesize the new image. DAMs using the Deep Boltzmann Machines(DBM) which is a mathematical model that has high order collections among input data for the face shaping structure and texture variation. This model also provides a bottom-up and top-down connection to effectively send updates between shape models and textural models In terms of shape models landmark points are used as an observation. Textural model, on the other hand, is more complex. It considers many factors ranging from facial expression, lighting to image resolution, for which can affect the values in texture model. For this reason, a shape free image is computed and wrapped the image under test. The goal is to remove the shape factors in modeling facial texture. DBM is used to create a shape free image. A binary RBM reveals the interactions within the hidden units in the higher layers. These hidden units have the power to increase the flexibility of the deep model. A higher layer is constructed to between the face shape and texture which encodes texture and shape information effectively and also serves as a connection layer. The benefits of using DAMs is that one can generate facial shape using textual data and vice-versa. Also when one of two data are not available the model is able to approximate the value. There exist an improvement to the DAM called the Robust DAM(RDAM) which provides a better reconstruction face results.