search

Deanplayerljx / tab-vcr

Pytorch implementation for our NeurIPS 2019 paper "TAB-VCR: Tags and Attributes based VCR Baselines" https://arxiv.org/abs/1910.14671
MIT License
18 stars 8 forks source link

the attributes and tag generation issues #6

Closed runzeer closed 4 years ago

runzeer commented 4 years ago

Great job!! But I have some questions about the attribute_feature and new_tag_feature generation. Could you share the codes about how to use the pretrained models in bottom-up-attention to generate the feature files?

Deanplayerljx commented 4 years ago

Hi,

I used the generate_tsv.py file in the bottom-up-attention repo repo to generate those features. For new tag feature, you can just change the input image to vcr images and run that code. For attribute_feature, you just need to use vcr's ground-truth bounding boxes as bounding box proposals and run the code.

Pefect96 commented 2 years ago

Great job!! But I have some questions about the attribute_feature and new_tag_feature generation. Could you share the codes about how to use the pretrained models in bottom-up-attention to generate the feature files?

I also need to extract the features and ask if you have generated them, thank you very much.