Silk is a natural fiber produced by silkworms (Bombyx Mori) and it has the unique property like strength, softness and lustrous. Silk is more popular all over the world because it’s used to produce luxurious cloths, medical sutures fashion accessories and other industrial applications. Silkworm rearing gives substantial annual income to the farmers and other related industries. Good quality silk depends on the selection of quality cocoons, until now the selection of high-quality cocoons is mostly performed as a manual process. This manual process is inadequate, highlighting the requirement for an automatic visual model to aid workers in the selection of silkworm cocoons. Recently, the deep learning (DL), models have achieved significant success and have become a one of the major models in defective cocoon identification. The current study was carried out for the classification of silkworm cocoons using residual graph convolutional neural network (ORE-GCN). The optimal process is carried out by the sandpiper optimization algorithm (SOA). The experimental analysis is performed on the real-time database and achieved a better accuracy of 98.1% and precision of 97.9%. This automatic model significantly minimizes the dependence on manual processes, ensuring robust classification of cocoons in silk production. Thus, the suggested model ORE-GCN provides a promising solution for the silk industry.