Questions about the MSR-VTT dataset

cshizhe / hgr_v2t

Code accompanying the paper "Fine-grained Video-Text Retrieval with Hierarchical Graph Reasoning".

MIT License

209 stars 21 forks source link

Questions about the MSR-VTT dataset #10

Closed Li-Zheng-94 closed 4 years ago

Li-Zheng-94 commented 4 years ago

Hi, Shizhe, thanks for your great work. I downloaded the MSR-VTT dataset you provided, and I have a question. I found that not every video corresponds to 20 captions. Some videos only correspond to less than 20 captions. I would like to ask if you specifically selected these captions and how to choose them?

cshizhe commented 4 years ago

There are redundant captions for some videos in MSRVTT dataset. I have removed these redundant captions.