search

farrell236 / RATCHET

RAdiological Text Captioning for Human Examined Thoraxes
MIT License
38 stars 9 forks source link

Used report section #4

Closed ChantalMP closed 1 year ago

ChantalMP commented 1 year ago

Hi! Thanks for your work and publishing your code. Which section of the report did you predict? The findings or impression section or both? Or the output of the create_section_files.py, which can be either of them depending on what's included in the report.

Thank you in advance!

farrell236 commented 1 year ago

Hi Chantal! In the paper, we predicted/evaluated on impression only as it was the default setting for create_section_files.py. We later retrained the model on findings (released as v2 weights in this commit tag).

Best wishes, Benjamin

ChantalMP commented 1 year ago

Thank you for your quick reply :)