Code for the ACL 2020 paper
- extract objects and annotations
- preprocessing/data_extraction.py
- train/test/dev splits
- preprocessing/data_split.py
- extract histograms
- preprocessing/histogram_colour/extract_histograms.py
- filter + resample data
- preprocessing/data_resampling.py
- train models
- models/bottomup/bottomup-train.py
- models/topdown/topdown-train.py
- models/earlyfusion/earlyfusion-train.py
- get model predictions
- models/bottomup/bottomup-predict.py
- models/topdown/topdown-predict.py
- models/earlyfusion/earlyfusion-predict.py
- evaluation
- evaluation/evaluation.ipynb
- evaluation/evaluation_unseen.ipynb
Data and models can be found here.