Since dry bean varieties have different qualities and economic values, their separation is of great importance in the field of agriculture. In recent years, the use of artificial intelligence-supported and image-based systems has become widespread for this process. This study aims to create a data set consisting of 14 classes in the detection of dry beans and to investigate the effectiveness of the hybrid structure of the extreme learning machine (ELM) model with GoogLeNet transfer learning on this dataset. At the same time, the salp swarm algorithm (SSA), which is one of the swarm intelligence algorithms, was used to test its applicability in ELM classifier by optimizing ELM parameters. The performance of these models was compared with ELM-based particle swarm optimization, harris hawks optimization, artificial bee colony, and traditional machine learning algorithms such as support vector machine and k-nearest neighbor. The suggested SSA-ELM model successfully classifies 14 different types of dry beans with a success rate of 91.43%. The comparable results demonstrate that the proposed hybrid model had better classification accuracy and performance metrics than traditional machine learning algorithms. In addition, it is seen that the use of image data, extraction of deep features, and classification with optimized ELM in the classification of dry beans have achieved comparable success in the literature.