docx4j doesn't have a page layout model, so it can't tell you a page count.
You can get an approximate page count by using FOP's page layout model. docx4j's PDF output now supports a "2 pass" generation:
- first pass calculates the page count(s)
- second pass generates the pdf
See https://github.com/plutext/docx4j/blob/master/src/main/java/org/docx4j/convert/out/fo/AbstractPlaceholderLookup.java and https://github.com/plutext/docx4j/blob/master/src/main/java/org/docx4j/convert/out/fo/ApacheFORenderer.java
So doing the first pass would give you (approximately) what you want. This uses org.apache.fop.apps.FormattingResults which records the number of pages in a page sequence, or in the document as a whole.
An alternative approach might be to use LibreOffice/OpenOffice (or Microsoft Word, for that matter).