Worker loading language traineddata progress 0

[IAndreaGiuseppe]

Describe the bug
Using a basic example code I'm unable to get an extracted text from an image.

Object { status: "loading tesseract core", progress: 0 }
Object { status: "loading tesseract core", progress: 1 }
Object { workerId: "Worker-0-ac418", status: "initializing tesseract", progress: 0 }
Object { workerId: "Worker-0-ac418", status: "initialized tesseract", progress: 1 }
Object { workerId: "Worker-0-ac418", status: "loading language traineddata", progress: 0 }

after this point nothing happen.

To Reproduce

<template>
    <div>
        <button v-on:click="recognize">recognize</button>
    </div>
</template>

<script>
import { createWorker } from "tesseract.js";

const worker = createWorker({
    logger: m => console.log(m)
});

export default {
    name: "ocr-reader",

    methods: {
        "recognize": function() {
            await worker.load();
            await worker.loadLanguage("eng");
            await worker.initialize("eng");
            await worker.initialize("eng");
            const {
                data: { text }
            } = await worker.recognize("http://localhost:8000/WEZK8.png");
            console.log(text);
            await worker.terminate();
        }
    }
};
</script>

simplest Vue component

Expected behavior
I expect to see a text message on console

Additional context
I'm doing a test on my localhost. I checked everything is correctly loaded. even traineddata file is correctly downloaded with status 200

[IAndreaGiuseppe]

Sooo, the problem is in Firefox, with chrome I can get the text extracted

[barryZZJ]

I'm using chrome and experiencing the same problem.

[jeromewu]

I have tried both Chrome and Firefox and it works perfectly.
@IAndreaGiuseppe may I know the version of Firefox?
@barryZZJ You might face some network issue, you can try this offline version to verify: https://github.com/jeromewu/tesseract.js-offline

[IAndreaGiuseppe]

@jeromewu I'm actually on FF 73.0.1 (64bit) on Windows

[jeromewu]

Hi @IAndreaGiuseppe, I have tried Firefox 74.0 (64bit) on Windows and it still works. Maybe you try Private Window in Firefox to avoid potential cache issue, and wait for a little more if your network is not fast.

[IAndreaGiuseppe]

Hi @jeromewu and thank you, I'm almost again on the subject and will be able to test this process again soon. Please don't close this issue.

[chungym]

I experience the same problem in firefox.
If langPath is set to a remote path such as 'https://tessdata.projectnaptha.com/4.0.0_fast', it works fine.
However, if langPath is set to a relative path inside the extension, it fails to load lang data.
No issue in chrome.

[TwoAbove]

I'm experiencing the same issue, but with any langPath. I tried the default one, https://tessdata.projectnaptha.com/4.0.0_fast, a relative path. All variants are stuck on loading language traineddata:

 { workerId: "Worker-0-2c495", status: "loading language traineddata", progress: 0, userJobId: "Job-1-6281d" }

Ubuntu, Firefox 96.0.2 (64-bit)

[TwoAbove]

The fix for me was to add

      cacheMethod: 'none'

[Balearica]

I think the only thing immediately actionable here is that the promise returned by worker.loadLanguage is neither resolved or rejected (no error message--it just gets "stuck"). Once that is resolved users can catch the error and try again with a different cacheMethod, and if there is an underlying bug with Tesseract.js we will have an error message to work off of.

It looks like this is the offending part--when a DOMException is encountered the promise is never rejected. I will edit such that all errors lead to a rejected promise.

tesseract.js/src/worker-script/index.js

Lines 147 to 153 in bce7cd8

	if (isWebWorker && err instanceof DOMException) {
	/*
	* For some reason google chrome throw DOMException in loadLang,
	* while other browser is OK, for now we ignore this exception
	* and hopefully to find the root cause one day.
	*/
	} else {

[Balearica]

I updated the master branch so now there should be no errors that do not lead to a message/promise rejection. This will be reflected in the next npm release (3.0.4). If anybody using version >=3.04 encounters either a non-resolving promise from worker.loadLanguage or an error due to an underlying bug they should open a new issue.