An Ensemble Learning-Based No Reference Qoe Model For User Generated Contents

Overview

This repository contains the source codes of our proposed QoE model for User-generated Content Videos described in the following publication:

Duc Nguyen, H. Tran and T. Thang, "An Ensemble Learning-Based No Reference Qoe Model For User Generated Contents," in 2021 IEEE International Conference on Multimedia & Expo Workshops (ICMEW), Shenzhen, China, 2021 pp. 1-6. doi: 10.1109/ICMEW53276.2021.9455959

To use our code, first following requirements.txt file to setup the enviroment.

Download dataset

Execute file dataset/download_video.sh to download the dataset that contains 7200 UGC videos. The total data size is ~35GB, thus it might take up to a day to download all the videos.

$ cd dataset
$ bash download_video.sh

Divide the download videos into dataset/train folder and dataset/test folder using information in train_label.csv and test_label.csv

Feature extraction

To extract features from the videos, run the following command from the terminal.

$ python3 feature_extractor.py

Extracted video features will be saved to dataset/train_feature.csv and dataset/test_feature.csv

Training

To train our model, run the following command from the terminal.

$ python3 train.py

The trained model will be saved to pretrained/model.pkl

Testing

Prepare a folder containing test videos, then run the following command.

$ python3 prediction.py /path/to/test/folder

Prediction results will be saved to a file named 'test_result.csv' in the current folder.