Recognition of Logo of Pirated Content Using Deep Learning-Based Regression Classification Algorithm

With the current state of online infrastructure, the phenomenon of content piracy is seeing rapid expansion and dissemination across several nations. The primary objective of internet infrastructure is to provide a platform for the transmission of authorized and legal content from the service provider to the recipient. Over the course of time, pirates have utilized the digital internet infrastructure system to replicate the original content and subsequently redistribute it using the same infrastructure. The utilization of visual analytics pertaining to the broadcast logo serves as a method for ascertaining the presence of stolen content. This study will focus on the development of a comprehensive collection of TV broadcast channel logos that are scalable and encompass a wide range of geographical locations and genres. Additionally, it will use publicly accessible statistics specifically pertaining to TV broadcast channel logos of Indian stations. A comprehensive collection of 450 television broadcast channel logos in many regional languages has been compiled, encompassing a range of categories such as sports, movies, kids and cartoons, entertainment, and more. To enhance the deep learning logo classification and expand the logo corpus, each logo undergoes exposure to a diverse range of data enrichment techniques. This study further investigates the identification of different logo classes using the use of the state-of-the-art object recognition approach known as YOLO. The experimental findings are recorded for several inference algorithms using different pixel contexts.