Open lvchw opened 8 months ago
Hi,
I am afraid you will have to write some custom code for this, as I didn't commit the code back then.
You can adjust this script by e.g. adjusting this line such that the model also outputs the bounding boxes together besides the beam search output. Then it's only a matter of the plotting the color coded region sentences and associated bounding boxes together.
Hi, thank you for your wonderful work!
I want to perform Visual Grounding on the Chest ImaGnome to visualize the results of report generation, but I failed to found any works to perform VG task on this dataset. I noticed that, in your paper, Appendix B.2 presents three excellent VG results. How are they produced? Is the code in the repository? If not, is it available? I would greatly appreciate it if you could help me.