The paper presents the result of using artificial intelligence to recognize geometric groups of parts made by metal forming. Five groups of axisymmetric parts produced using cold axial rotary forging have been selected as the object of research. To recognize the details belonging to these groups, a convolutional neural network ResNet50V2 based on TensorFlow and the Keras API was synthesized. After training the network, the recognition rate for the selected groups was at least 90%. The proposed approach to determining the geometric group of parts allows us to move away from the subjective assessment of the geometry of parts, and to fully digitalize the design process with the implementation of the advantages of the group manufacturing method.