As of 2019, 720,000 hours of video is uploaded on YouTube alone, and 90% of people say that they discover new brands and products on YouTube. Moreover, due to the ongoing pandemic, the consumption of video content has increased exponentially. The MediaEval 2018 proposed a task of predicting the memorability of these videos which would help content creators to make more impactful content. In this study, I have designed a model using provided Video and Image features to predict both the short-term and long-term memorability scores. Finally, the results showed that aweighted average ensemble of different models gives better results than a single model trained on a feature.
- Code --> Contains the full code
- Conference Paper --> Contains a conference style paper which has all the details of the project.