Joey Yang

I think it originates from here: https://github.com/ChenRocks/BUTD-UNITER-NLVR2/blob/master/data/genome/1600-400-20/objects_vocab.txt (0 for background)

The default tokenizer in prepro.py is 'bert-base-cased'.

Hi, I found the link was broken, could you please update the link? Thanks!

You can reach me via email.

Hi, same question here, have you managed to generate features/label prediction for test images yourself? I found that the author provide another github repo to generate features: https://github.com/microsoft/scene_graph_benchmark but I...

I think you can find this one in https://github.com/alfworld/alfworld/blob/master/alfworld/gen/constants.py#L1206 since they are using the pretrained detector from this repo.