The implementation of Group technology (GT) involves the use of a Classification and Coding (CC) system and/or a direct analysis of the production methods. Classification and Coding is a laborious and time-consuming activity, and has been one of the main impediments for the implementation of GT. A considerable time is also spent in the selection of a coding system and training. Thus the need for computer-assisted procedures in this area can be clearly understood. This paper first gives a review of the approaches that have been used for automation of the activity of classification and coding. Next it presents a part feature recognition system that eliminates the human effort of translating part definition data (from a CAD drawing), which is geometry based into a scaled bitmap pattern representation of part drawings which serves as input for a neural network GT code generator. The neural network, designed with back-propagation, is used to generate the part geometry related digits of the VUOSO code from the transformed bitmaps of the CAD drawings. This interface via a CAD primitives processor not only reduces time for classification and coding, but also eliminates any need to change the neural network because of design changes in CAD. In addition to making the system less error prone, further recoding of parts are also easier because of the adaptability of neural networks. The effort thus shows the feasibility of CAD to a CC system and automation of the process of GT code generation.