NOT able to reproduce spearman and kendall tau as reported in the paper. - Githubissues

TIBHannover / MSVA

Deep learning model for supervised video summarization called Multi Source Visual Attention (MSVA)

MIT License

41 stars 18 forks source link

NOT able to reproduce spearman and kendall tau as reported in the paper. #15

Open rodosingh opened 1 year ago

rodosingh commented 1 year ago

Hello,

While reproducing the results, I'm unable to get the scores (Spearman and Kendall Tau) that you guys reported in the paper. It's deviating a lot from the reported score.

Can you please provide the actual code for computing it?

As per lines 430 and 431 (shown below) https://github.com/TIBHannover/MSVA/blob/dad26a69460a08ef82f23a26f34c13c6e6fc3c34/train.py#L430-L431 We tried all different combinations but were unable to reach near what you have mentioned. Please guide us!

Thanks in advance 😃 .