Enhanced ML Based Content-Based Image Retrieval for Mobile Devices

Image retrieval systems based on content can be developed for mobile applications that provide users with a seamless and efficient way to search for images based on their content. The development of CBIR systems for mobile applications presents unique challenges and opportunities. Mobile devices typically have limited computational resources, storage capacity, and network bandwidth compared to traditional computing platforms. Therefore, CBIR systems designed for mobile applications must be lightweight, efficient, and optimized to run smoothly on these devices. The Adaboost algorithm progressively learns weak classifiers and aggregates them to construct a robust final classifier. Subsequently, a weak learner is introduced, and the data undergoes reweighting. Misclassified test images are assigned greater importance while correctly classified ones are downweighted. Consequently, subsequent weak learners prioritize the query images previously misclassified by their predecessors. The average precision is increased in further iterations to 95.6% from the initial iteration of 94.9%. Average recall shows a little more improvement from the previous set by increasing from round 1 with 94.1%, and increasing to 95.9% towards the last round of iteration. The F-Measure keeps fairly stabilized at 95.1478 after the first round. The time taken by every iteration reduces slightly from 3.