Is there any reason PMS options are restricted?

[rony007dennis]

I wanted to use the full set of PSM options especially 1. Why is there only a limited set of parameters available?

n8n-nodes-tesseractjs/nodes/TesseractNode/TesseractNode.node.ts

Line 313 in 031c1c6

const psm = this.getNodeParameter('options.psm', 0, 'SINGLE_BLOCK') as 'SINGLE_BLOCK' | 'SINGLE_COLUMN' | 'SINGLE_LINE' | 'SINGLE_WORD' | 'SPARSE_TEXT';

Wanted to use option 3 from here...
tesseract-ocr/tesseract#1651 (comment)

Full list explained here too:
https://pyimagesearch.com/2021/11/15/tesseract-page-segmentation-modes-psms-explained-how-to-improve-your-ocr-accuracy/

[jreyesr]

Hello! I can't quite remember, but I think the rationale was:

• PSM 0 doesn't return any OCR information (no text is parsed, just some information about it), and so should be exposed, if anything, as an additional operation, not in the same bag as the detection modes
• PSMs 1 and 2, at least according to https://pyimagesearch.com/2021/11/15/tesseract-page-segmentation-modes-psms-explained-how-to-improve-your-ocr-accuracy/, aren't implemented, what they do is unclear or they don't work well: "in all my experiments, I could not find a situation where --psm 1 obtained a result that the other PSMs could not" and The --psm 2 mode is not implemented in Tesseract"
• PSM 5, at least in the example given in the pyImageSearch post above, can be emulated by rotating the image beforehand and then using PSM 4, which is available
• PSM 9 seems quite exotic (words in a circle), so I skipped it to keep the options dropdown cleaner
• PSM 10 can probably be emulated by parsing as a single word, that just happens to have a single char
• PSM 12 doesn't seem to work well (same article, "That said, I had a lot of problems getting this mode to work properly and could not find a practical example where the results meaningfully differed from --psm 11.")
• PSM 13 is the last resort and shouldn't probably be used much, "Using --psm 13 can be a bit of a hack at times so try exhausting other page segmentation modes first."

That leaves PSMs 4, 6, 7, 8 and 11, which are the five available modes, and PSM 3 (Auto, actually the default one), which I can't for the life of me remember why I didn't make the default. I believe it was because it should always behave as one of the available modes, and so it should be possible to manually set the node to parse blocks, columns, lines or words.

So, in short: it was to keep the list of options short and more manageable, especially given that some modes don't seem to work and are therefore liable to cause confusion, and the node biases slightly towards non-expert users so there's an incentive to keep things short and simple (expert users are more likely to already have their complex Jupyter notebooks or whatever instead of using N8N)

Are you having issues with the available modes? The default mode SINGLE_BLOCK should work much like mode 3, it should return full paragraphs. Or I could add PSM 3 and make it the default, to match the Tesseract CLI, I just don't have any other changes to bundle this with. Any other modes that you'd be interested in?

[rony007dennis]

I wanted to deal with the rotated images. Which is autodetected in mode 1 it seems. Otherwise, I would need to do some fiddling around rotating it. This is doable, but it will cause an unnecessary complex step in between.
Havent tried the package as it is with this mode to verify but this post says it can.

[jreyesr]

Got it. I should be able to make the change in a day or two.

---

In the meantime, if you need this now or want to test things early, you can (if you installed N8N using NPM, and you run it with n8n start. This doesn't work with Docker) manually edit the file ~/.n8n/nodes/node_modules/n8n-nodes-tesseractjs/dist/nodes/TesseractNode/TesseractNode.node.js and add the following two files around line 160:

                                {name: 'Auto OSD', value: "AUTO_OSD", description: ""},
                                {name: "Auto (3)", value:"AUTO", description: ""},

like this, after the opening bracket [

Then, stop and restart your N8N instance, and then you should see two new options in the PSM dropdown:

Could you please try both and let me know if you find any difference between the "Auto OSD" and "Auto (3)" options? From my tests with a manually rotated image, I couldn't see any differences, and I've found two or three posts that seem to hint that, unlike the 2018 comment in the Tesseract repo, PSM 3 works well enough with rotated images, and so PSM 1 isn't necessary. If that is true, I'd prefer to only add the Auto mode, which also happens to be the default one in the Tesseract CLI, instead of two new modes where one doesn't work any differently from the other.

[rony007dennis]

Not that urgent. I have decided to use the python code directly rather than n8n by the way as I needed some more granular control. Thanks, Rony

…

On Mon, 24 Mar 2025 at 1:49 am, jreyesr ***@***.***> wrote: Got it. I should be able to make the change in a day or two. ------------------------------ In the meantime, if you need this now or want to test things early, you can (if you installed N8N using NPM, and you run it with n8n start. This doesn't work with Docker) manually edit the file ~/.n8n/nodes/node_modules/n8n-nodes-tesseractjs/dist/nodes/TesseractNode/TesseractNode.node.js and add the following two files around line 160: {name: 'Auto OSD', value: "AUTO_OSD", description: ""}, {name: "Auto (3)", value:"AUTO", description: ""}, like this, *after* the opening bracket [ image.png (view on web) <https://github.com/user-attachments/assets/ed219a04-dd28-4fed-9550-7a1c32bafeb3> Then, stop and restart your N8N instance, and then you should see two new options in the PSM dropdown: image.png (view on web) <https://github.com/user-attachments/assets/618bedb7-c85f-4e1d-8dec-14cb0d2d5931> Could you please try both and let me know if you find any difference between the "Auto OSD" and "Auto (3)" options? From my tests with a manually rotated image, I couldn't see any differences, and I've found two or three posts that seem to hint that, unlike the 2018 comment in the Tesseract repo, PSM 3 works well enough with rotated images, and so PSM 1 isn't necessary. If that is true, I'd prefer to only add the Auto mode, which also happens to be the default one in the Tesseract CLI, instead of two new modes where one doesn't work any differently from the other. — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADVM4B3V4B5XTXOJ5MISHB32V5QBZAVCNFSM6AAAAABZSMWYICVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONBWGY3TCNZRGM> . You are receiving this because you authored the thread.Message ID: ***@***.***> [image: jreyesr]*jreyesr* left a comment (jreyesr/n8n-nodes-tesseractjs#2) <#2 (comment)> Got it. I should be able to make the change in a day or two. ------------------------------ In the meantime, if you need this now or want to test things early, you can (if you installed N8N using NPM, and you run it with n8n start. This doesn't work with Docker) manually edit the file ~/.n8n/nodes/node_modules/n8n-nodes-tesseractjs/dist/nodes/TesseractNode/TesseractNode.node.js and add the following two files around line 160: {name: 'Auto OSD', value: "AUTO_OSD", description: ""}, {name: "Auto (3)", value:"AUTO", description: ""}, like this, *after* the opening bracket [ image.png (view on web) <https://github.com/user-attachments/assets/ed219a04-dd28-4fed-9550-7a1c32bafeb3> Then, stop and restart your N8N instance, and then you should see two new options in the PSM dropdown: image.png (view on web) <https://github.com/user-attachments/assets/618bedb7-c85f-4e1d-8dec-14cb0d2d5931> Could you please try both and let me know if you find any difference between the "Auto OSD" and "Auto (3)" options? From my tests with a manually rotated image, I couldn't see any differences, and I've found two or three posts that seem to hint that, unlike the 2018 comment in the Tesseract repo, PSM 3 works well enough with rotated images, and so PSM 1 isn't necessary. If that is true, I'd prefer to only add the Auto mode, which also happens to be the default one in the Tesseract CLI, instead of two new modes where one doesn't work any differently from the other. — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADVM4B3V4B5XTXOJ5MISHB32V5QBZAVCNFSM6AAAAABZSMWYICVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONBWGY3TCNZRGM> . You are receiving this because you authored the thread.Message ID: ***@***.***>