Semantic Analysis and Machine Learning Techniques for Enhancing Content-Based Image Retrieval

Multimedia data has grown in volume and complication due to the explosion of computing innovations and the introduction of the Internet. The client can submit a request to the content-based image retrieval (CBIR) structures, which allows it to obtain the needed photo from the image collection. This platform has been designed as an effective image-retrieving instrument. Unfortunately, there are several issues with the typical relevant response of CBIR that would reduce the platform's efficiency, including imbalances in the training collection and categorization problems, inadequate learning collection problems, and user problems with constrained data. This paper suggests using a machine learning-driven (ML) semantic analysis for CBIR with local neighbor patterns (LNP).To increase the efficacy of this platform, this study employed 3 different kinds of datasets: texture, hue, and face datasets. This study employed the LNP approach in conjunction with ML methods to boost the accuracy of this platform. The performance evaluation demonstrates that the application of ML methods in conjunction with regional patterns enhances the mean accuracy from 36.2 to 85.6% when using LNP with cubic the Support Vector Machine (SVM), from 82.5 to 99.5% when using LNP with acceptable kernel neural network (KNN), and from 56.6 to 95% when using LNP with ensemble subdomains discriminatory.