Development and Evaluation of improved Image Retrieval and Classification method using YolovV8 and efficient Deep Network Model

A novel image retrieval and classification system has been developed to overcome the limitations of conventional image retrieval methods, facilitating efficient image searching. During the training phase, images are sourced from a publicly available dataset and processed through a segmentation step, utilizing YOLOV8 for precise segmentation. The segmented images are then classified using a Feature Pyramid Network-based Residual Attention Network (FPN-RAN).In the testing phase, user-provided query images are processed to retrieve the most relevant matches from the database. Deep features of the query images are extracted using the FPN-RAN model and compared to the pre-stored trained image features in the database. A multi-similarity function is employed to evaluate the similarity between the query and database images. Several similarity measures are integrated within the multi-similarity function to ensure accurate and prompt retrieval of suitable images based on the similarity metrics. The model’s effectiveness is validated by comparing its performance against previous image retrieval techniques using multiple simulated evaluation metrics. These empirical results highlight the superiority of the proposed system.