mrraghav / media-memorability

MediaEval challenge 2019 - to predict the memorability of the Videos

This project is an attempt towards Media Memorability section of mediaeval2019 in which we predict the memorability of a video to the viewers.
Link (2019) : http://www.multimediaeval.org/mediaeval2019/memorability/

Since last decade, MediaEval offers yearly challenges in the forms of shared tasks.

Website (2010-2019) : http://www.multimediaeval.org/
Website (2020-) : https://multimediaeval.github.io/

Video memorability has various applications. In this project, features extracted by a different team have been utlized to predict the short- and long-term memorability of the videos.

Features used:

HMP
C3D
C3D with HMP
captions using count vectorizer
captions using tf-idf
captions with HMP
captions with C3D

Regression techniques used:

Linear Regression
KNN
Decision tree
Random forest
VotingRegressor
AdaBoostRegressor
GradientBoostingRegressor
Neural Network

After analyzing the spearman's score for all of them, the best regressor was used for the test set.