Content Based Image Retrieval Using Multi-deep Learning Models and K-Nearest Neighbor Approaches

Content-based Image Retrieval (CBIR) is a pivotal process for querying images and retrieving relevant results. This research carries a significant impact across diverse applications in our daily lives. Despite the myriad of proposed approaches, this problem continues to pose a formidable challenge. In this study, we advance the quality of image search through the integration of K-Nearest Neighbor with multiple deep neural networks in CBIR. Initially, crucial features are extracted utilizing pre-trained CNN models. Subsequently, we gauge the similarity between feature vectors via the K-Nearest Neighbors algorithm. Our experimentation is conducted on the Oxford-IIIT Pet Image Dataset, and the model's efficacy is assessed across diverse CBIR metrics. The results of our experiments substantiate the exceptional performance of our proposed system.