hudaalamri/dstc7-audio-visual-scene-aware-dialog-avsd-challenge

dstc7-audio-visual-scene-aware-dialog-avsd-challenge's Issues

can't reproduce results

Hey, I tried to reproduce the github results. It seem some files are missing (Maybe you call it train/test/valid.list, a list of split movie indices.) I've made it from the json file and it seems to work, but I keep getting lower results. Moreover, i3d-rgb works better than i3d-rgb + i3d_flow.

To make it simpler, is it possible to share the weights of a trained model, and charades2text_i3d_rgb_features.pkl and charades2text_i3d_flow_features.pkl files?
Thanks

Deadline

Hi,

I'm a bit confused regarding when can make submissions and what's the deadline for the challenge.
Link: http://workshop.colips.org/dstc7/dates.html

Can we make submissions until the end of development phase or evaluation phase?

Thanks in advance,
Raghav

Download files script/Replace file storage provider

Hey, I find Google Drive a bad choice to share files, because it's not trivial to use wget/curl with.
It will be much simpler if you can provide a script to download necessary files, or alternatively replace google drive with other storage provider (Dropbox?).

Problem running the feature extraction

I've been trying to run this code for a while now.
Some of the libraries used have been updated (namely the ones used in the i3f feature extraction). Can you let me know which version of tensorflow and sonnet were used? I feel there are some compatibility problems with the latests versions of both libraries.