Bio-Inspired Optimization Algorithm on Cloud based Image Retrieval System using Deep Features

With the tremendous improvement of cloud computing (CC) technique, increasing count of users choosing to outsource image data to the clouds. Cloud resource provider decreases the storing problem on local hardware by outsourcing massive image database to cloud server and exploiting the cloud computation ability to image processing application. Learning effective feature representation and similarity measures are essential for the performance of content-based image retrieval (CBIR). Despite wide-ranging research efforts for decades, it remains a challenge that significantly hinders the success of real-time CBIR system. Since the trail and error hyperparameter selection is a tedious process, metaheuristic optimization can be used. This paper introduces an Evolutionary Optimization Algorithm for Cloud Based Image Retrieval System (EOA-CIRS) technique. The presented EOA-CIRS technique derives a new CBIR model for retrieving highly related images. To obtain this, the EOA-CIRS technique initially employs MobileNetv3 model as a feature extractor, which derives features from the query and database images. Next, the bamboo forest growth optimizer (BFGO) approach was applied as a hyperparameter optimizer to appropriately tune the hyperparameters of the MobileNetv3 approach. Finally, Euclidean distance based similarity measurement is utilized for retrieving the related images. The experimental validation of the EOA-CIRS model on Corel10K dataset demonstrates the effectual efficacy of the EOA-CIRS model over other recent approaches.