Xuanyu Zhou

I think the online demo used stanford_3.3.1 default for DEPENDENDY_STANFORD view. Could that be the reason why the results are different from 3.8.0? @mssammon @nitishgupta

Thanks, we have realized this and will work on something new. Meanwhile, if you want to work with other transformers, you will need to modify pattern_extraction.py, which contains the entire...

Hi, thanks for your interest in the work 1. That's right, the model cannot recover all the masks, nor the human. The goal here is to give the model a...

What are the output values in your version? I will try to find how I computed it exactly. Nonetheless, these are hard-coded distributions and they should not "cheat" anything. Small...

It seems that the standard deviations are different, where mine is much smaller so the peak value has a larger probability (0.585), and converges to 0 much faster (a 2-distance...

I see the issue now. I will try to find what exactly happened during that computation and get back to you. Right now I suspect that some kind of scaling...

Okay as it turned out, I used a [x / sum(X)] normalization instead of a softmax. The ratio in the final probability vectors are 1-1 proportional to those in the...

Yes, here is a quick re-implementation, please let me know if you spot anything out of the ordinary: ``` from scipy.stats import norm convert_map = { "seconds": 1.0, "minutes": 60.0,...

Sorry about that, we are migrating some of our servers which caused the issue. I will update the download link. Meanwhile, please use this link: https://drive.google.com/drive/folders/1fD6WfCEPQICGPhxqlwuVmf8uOot-jQq8?usp=sharing Please let me know...

I just updated data.zip in the above link so that it contains all datasets. Please re-download if you did so before this note. Sorry for the trouble.