Junbin Xiao

icon indicating a website link https://doc-doc.github.io/cv/

icon company National University of Singapore icon location Singapore

Results 55 comments of Junbin Xiao

Some videos are not working

Hi, please try [VLC Player](https://www.videolan.org/).

How to get Relation annotations.

Hi, please find the annotations [here](https://xdshang.github.io/docs/vidor.html). For some missing videos , please refer to [here](https://drive.google.com/file/d/1l1VdNgQiltaJjSmg9QkcxDWbwXXwnIMH/view?usp=share_link)

Unknown definition of "qid" and "type" in annotations

Hi, thanks for the interest. `qid' gives the question id within a video. So video_qid identifies a sample of the whole dataset. There is no binary question in the multi-choice...

Unknown definition of "qid" and "type" in annotations

Yes.

Unknown definition of "qid" and "type" in annotations

Hi, please refer to the evaluation file eval_mc.py..

The visual features for STVQA

Sorry that I am recently too busy to go back to this work and manage the feature (it's large). I will release it after next week.

The visual features for STVQA

Hi, the spatial features are available for downloading.

The visual features for STVQA

Sorry that I have removed the feature for the limited cloud storage as also the method is out-of-date. If you really want the feature, please use the code provided by...

Can you provide the BERT features of the candidate answers?

Thanks for your interest. As we finetune BERT for multi-choice QA, each answer is appended to the question as a global sentence. If you want the separated candidate feature, you...

Can you provide the BERT features of the candidate answers?

Hi, currently, we are not going to release this part of code, but you can refer to this [file](https://github.com/LuoweiZhou/pytorch-pretrained-BERT/blob/master/examples/run_swag.py) for details.