Figures and tables in the back / annex section ignored

[de-code]

This is related to #698

Some documents have main figures and supplementary figures.
If in those cases, the segmentation model labels the supplementary figures as annex,
then the content is passed separately to the fulltext model.
If the fulltext then correctly labels it as figure, then the figures from the annex are not included in the output.

[de-code]

This seems to be due to FullTextParser processing figures and tables from the body only.

[lfoppiano]

@de-code do you have a Pdf for testing?

[de-code]

One example is DOI 10.1101/306803 or 306803v1 (from the bioRxiv 10k validation dataset).
It has "Extended Data Figure 1" etc.
I haven't tested whether they are going to get extracted well with the default models.

[lfoppiano]

Hi @de-code (long time 😄), I stumbled upon this issue now, indeed it seems the tables in the back are not processed.
According to my understanding of the guidelines, they should be output in the annex, as they are annotated for the segmentation model.
The PR #963 is slowly gathering priority 😸