/VideoQADenseCapFrameGate-ACL2020

Code for ACL 2020 paper "Dense-Caption Matching and Frame-Selection Gating for Temporal Localization in VideoQA." Hyounghun Kim, Zineng Tang, Mohit Bansal.

Primary LanguagePythonMIT LicenseMIT

VideoQADenseCapFrameGate-ACL2020

Code for ACL 2020 paper "Dense-Caption Matching and Frame-Selection Gating for Temporal Localization in VideoQA." Hyounghun Kim, Zineng Tang, Mohit Bansal.

Prerequisites

  • Python 2.7
  • PyTorch 1.1 or Up
  • Install all the required packages from requirements.txt file.
pip install -r requirements.txt

Datasets:

Please downlaod basic data from the TVQA+ web page here.
Please download pre-processed data from here.

Usage

To train the model:

bash main_run.sh

Best pre-trained model can be downloaded from here.

Acknowledgments

Base code is from the TVQA+ code repository.