Deep Learning based Telugu Video Text Detection using Video Coding Over Digital Transmission

With the development of smart devices and high-speed internet, demands for the ever-increasing number of videos. Comparing conventional machine learning methods with in-depth Reading methods. The two independent steps used to decipher text is; the discovery of the Candidate text region and the division into categories. Then, the text texts received were. distributed through a few layers of CNN, which first produced a conversion feature map as well as Divide candidate circuits into text or non-text classes. Explored the proposed programs. On the data got from various Telugu videos. The data sets have been collected by collecting. videos from various Telugu channels. Text from the video provides useful information, which can create a default video index and retrieval system and integrated heuristic and in-depth learning-based approach Neural Network (CNN) to automatically output text to video. Two independent steps were used to extract the existing text; the discovery of the Candidate text region and the division into categories. In the first step, the circles are rectangular. It identified textual content using heuristics, which included morphology analysis and geometric issues. Then, the text of the accepted candidate text has transferred to a few layers of CNN, which first produced a convolutional feature map and then split the candidate regions into text or non-text classes. Video collection organized on the website is different. channels Telugu. 70 percent of the data have been used for network training while 30 percent Confirmed. Experiments have shown that our proposed approach has reached modernization functionality in our database.