Mathematical modelling and statistical analysis in designing deep learning-based shot boundary detection and aesthetic assessment of videos

Abstract

The main goal of this study is to create a strong deep learning-based system for finding shot boundaries and judging the quality of movies by combining mathematical models and statistical analysis. Aesthetic evaluation helps people understand visual material better, and shot border recognition is an important part of video analysis. We come up with a new approach that uses convolutional neural networks (CNNs) to improve the accuracy of shot border recognition by extracting features and using mathematical modeling. The model’s performance is improved through statistical analysis, which makes sure that changes between shots are correctly identified. Proposed method also includes judging how something looks by combining deep learning methods with mathematical models that understand how things look. This two-in-one model gives a full picture of video material, which helps with both content selection and improving the user experience. We test our method on a variety of video datasets and show that it works better than other methods. The way our deep learning approach uses mathematical models and statistical analysis together not only raises the bar for finding shot boundaries, but it also helps the new field of judging the aesthetic quality of video material. This study creates a useful and flexible tool for analyzing videos, which opens the door to better uses in fun, content creation, and multimedia. The proposed method is implemented MATLAB and the performances evaluated based on performance metrics such as such as accuracy, precision, sensitivity, specificity and recall. And the accuracy of the proposed system is obtained as 93.12%. And to assure that the proposed system functions effectively in aesthetic assessment of video a contrast is made with the conventional technique such as Hand-Crafted Feature (HCF) technique, Neural Network (NN) and Support Vector Machine (SVM).