Content-based image retrieval (CBIR) has gained significant attention in recent years due to the exponential growth of digital images and the need for efficient image search and retrieval systems. Deep learning, with its ability to automatically learn hierarchical representations from raw data, has emerged as a powerful tool for CBIR tasks. This paper proposes a CBIR system by merging DarkNet-19 and DarkNet-53 information for logo image retrieval. Experimentation is done using the Flickr Logos-47 dataset. The results revealed a significant enhancement in the retrieval results in terms of precision as compared with the standard image retrieval methods and the Deep Convolutional Neural Network(DCNN).