Most manufacturing industries are facing changes due to the increasing competition in the global market. Regarding this current situation, manufacturing firms that offer mass customization usually have a competitive edge over counterparts offering generic products. However, the Mass Production-to-Mass Customization (MP-to-MC) transition has brought upon unprecedented challenges to many Small and Medium Enterprises (SMEs). One of the challenges is to identify customized products in harsh production environments. The reason behind this is that identification tags such as barcodes, Quick Response (QR) codes and Radio Frequency Identification (RFID) cannot function in special production processes like heating and dissolution. It is therefore of prime importance to find a solution by devising a product identification method without making use of marks or tags. In the paper, a novel method that using computer vision to identify the customized products in mass customization and a hybrid Convolutional Neural Network (CNN) model are proposed. To illustrate the efficacy of the proposed method, a case study in a shoe-manufacturing company was reported. The results yielded demonstrated that the proposed method is an efficient and economical solution.