allenai / vila

Incorporating VIsual LAyout Structures for Scientific Text Classification
Apache License 2.0
167 stars 17 forks source link

Better e2e parsing #34

Closed lolipopshock closed 1 year ago

lolipopshock commented 1 year ago

Before this PR, all the visual equation detection results are only saved as images but not in the structure.csv file. We fix this by pairing the text-based equation token classification results and the visual equation detections in a way similar to figure and caption matching and save it in the structure.csv file.

We also add the vila paper to the tests for #33.