Unifying the Video and Question Attentions for Open-Ended Video Question Answering

Table of Contents

<!--ts-->

• Introduction
• Datasets
• Methods
  • Compared Algorithms
  • Results
• Dependency
• Usage
• Reference
• License

<!--te-->

Introduction

videoqa is the dataset and the algorithms used in Unifying the Video and Question Attentions for Open-Ended Video Question Answering

Datasets

• file_map: contains the Tumblr urls of the videos
• QA: contains the question-answer pairs
• Split: contains the dataset split in the paper

Methods

Compared Algorithms

• [E-SA] (https://www.aaai.org/ocs/index.php/AAAI/AAAI17/paper/viewFile/14906/14319)
• [SS-VQA] (https://www.aaai.org/ocs/index.php/AAAI/AAAI17/paper/viewFile/14906/14319)
• Mean-VQA: a designed baseline where imageQA is performed on each frame

Results

• Question: What is a boy combing his hair with?
• Groundtruth: with his fingers
• Prediction: with his hands

• Question: What runs up a fence?
• Groundtruth: a cat
• Prediction: a cat

• Question: What is a young girl in a car adjusting?
• Groundtruth: her dark glasses
• Prediction: her hair

Dependency

• Theano
• Blocks
• Python >= 3.4

Usage

python main.py

Reference

If you use the code or our dataset, please cite our paper

@article{xue2017unifying,

title={Unifying the Video and Question Attentions for Open-Ended Video Question Answering},

author={Xue, Hongyang and Zhao, Zhou and Cai, Deng},

journal={IEEE Transactions on Image Processing},

year={2017},

publisher={IEEE}

}