A Hybrid Machine Learning Model for Image Deduplication

In today's cloud computing landscape, the exponential growth of data generation presents new challenges for service delivery. As individuals increasingly create and store various types of content such as text, audio, images and videos, the demand for efficient data storage solutions becomes paramount. Cloud storage offers a cost-effective way to manage this data, but it also raises concerns about data duplication, particularly in a world where selfies and similar images are prevalent. To address these challenges, this paper proposes an image deduplication system designed to optimize storage by identifying and eliminating similar images before they are uploaded to the cloud using Resnet50, efficient net and mobile net with the clustering algorithms as Triplet and pairwise algorithm. By assessing the likeness of newly uploaded images against those already stored, it can effectively reduce redundancy, minimize storage costs by ensuring a more efficient use of cloud resources with 91% accuracy.