[
https://issues.apache.org/jira/browse/TIKA-4666?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18059181#comment-18059181
]
ASF GitHub Bot commented on TIKA-4666:
--------------------------------------
tballison merged PR #2614:
URL: https://github.com/apache/tika/pull/2614
> Add VLM/modern OCR options parsers in 4.x
> -----------------------------------------
>
> Key: TIKA-4666
> URL: https://issues.apache.org/jira/browse/TIKA-4666
> Project: Tika
> Issue Type: New Feature
> Reporter: Tim Allison
> Priority: Major
>
> Along the lines of TIKA-4665, it would be great if we could integrate with
> modern vlms. Ideally, these would call out to vlm apis, but we could include
> examples of running qwen or jinaai locally in a python server.
> Again, as with our old deeplearning4j module, I think we should offer
> examples of these kinds of integrations whether or not they get picked up in
> production.
> Modern VLMs can be a swap in for tesseract ocr depending on the prompt. Or,
> they can be used to label images... user chooses what the prompt will be.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
