NORMA eResearch @NCI Library

Categorization of Audio/Video Content using Spectrogram-based CNN

Rajput, Nishant (2019) Categorization of Audio/Video Content using Spectrogram-based CNN. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
Preview
PDF (Master of Science)
Download (1MB) | Preview

Abstract

Understanding the content in the videos or audios has been an important task since a long time now. By developing the understanding of content the world could actually become a safer place for example if the hatred videos could be blocked before it spreads or the adults content can be prevented to be shared with the children. Imagine if the systems were smart enough to stop the hatred spread during Christchurch shootings, then maybe christchurch massacre wouldn't have happened as the main motive behind the shootings was to spread the hatred. This research steps towards the identification of content in the audio and videos using the embedded audio present in the audio. In this research a state-of-the-art audio CNN is developed which could even run on the basic machine and do not necessarily require high end devices to run on the cost of loosing the accuracy by 3-4%. This model can run up to the accuracy of 94.88% and this model is able to classify the content even in the presence of the background noise though the performance get hampered a bit by the induction of noise but it is still feasible enough to run and obtain output from the model.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QA Mathematics > Computer software
T Technology > T Technology (General) > Information Technology > Computer software
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Caoimhe Ní Mhaicín
Date Deposited: 14 Oct 2019 08:53
Last Modified: 14 Oct 2019 08:53
URI: https://norma.ncirl.ie/id/eprint/3859

Actions (login required)

View Item View Item