search

YunseokJANG / tgif-qa

Repository for our CVPR 2017 and IJCV: TGIF-QA
https://arxiv.org/abs/1704.04497
171 stars 31 forks source link

Why the result is different from paper? #16

Closed Apm5 closed 4 years ago

Apm5 commented 4 years ago

The results in the READEME.md are quite different from those in the paper. Why?

What optimizations have been made in the code?

YunseokJANG commented 4 years ago

Hi @Apm5, First of all, thank you for your interest in our work. I checked the performance of our up-to-date CVPR work link, and the number is matched with this. As we mentioned in the last paragraph of Section 1, we've increased the size of our dataset, and that was the reason of the updated performance.

Regards,

Apm5 commented 4 years ago

@YunseokJANG Thank you for reply. I'm following this link, which is Google scholar search result.

YunseokJANG commented 4 years ago

@Apm5 Thanks for mentioning this. I tried to change the link in my Google Scholar profile, to make less confusion, but I failed. Sorry.

By the way, at the same time, I would like to recommend to check arXiv if you want to check what is the most recent one. Because the camera-ready version cannot be changed (by definition), people update their results on arXiv if there is any.