26hzhang / VSLNet

Span-based Localizing Network for Natural Language Video Localization (ACL 2020)
MIT License
101 stars 17 forks source link

about visual features for tacos dataset #14

Closed Aarontncl closed 1 year ago

Aarontncl commented 2 years ago

Here I'm wondering what is the difference between the 2D-TAN tacos features and the TALL tacos features, i.e. what preprocessing does 2D-TAN do to modify the original dataset? It seems 2D-TAN is using the same split as TALL. And I find there are two types of extracted features provided in the box drive, so I'm also a little confused about the difference between them.

Thanks a lot!