For anyone else who stumbles across this: you need to supply an errorHandler in the options dict when you initiate Tesseract. See below PR:

#368

Something like:

import { createWorker } from 'tesseract.js';

const worker = createWorker({
  logger: m => console.log(m),
  errorHandler: e => myErrorHandler(e)
})

I am attempting to ocr a batch of images, but with this kind of error handling there is no way to reuse a worker and just skip over images that throw an error. It would just be nice if the recognize function could throw the error so you can handle it on a per job bias.

I figured out how to handle errors when using one worker to go through a large batch. I will leave my last comment here because it might be helpful to someone to find this, but I realized after quite a bit of digging that if you pass an error handler to the create worker function, even if it does nothing, it will allow the error to be thrown in the promise for the recognize function. But because this error causes the work thread to stop working, you still do have to do something to fix that. My solution was just to create a new thread when the promise fails.

My solution:

async function setupWorker() {
  const worker = createWorker({
    // eslint-disable-next-line @typescript-eslint/no-empty-function
    errorHandler: () => { }
  });

  await worker.load();
  await worker.loadLanguage('eng');
  await worker.initialize('eng');

  return worker;
}

async function cleanUpWorker(worker: Worker) {
  await worker.terminate();
}

async function main() {
  const imgs = ['./img1.png', './img2.png', './img3.png']
  let worker = await setupWorker();
  for (const img of imgs) {
    for (let i = 0; i <= maxRetries; i++) {
      try {
        const result = await worker.recognize(img);

        // TODO: upload result to database

        break;
      } catch (err) {
        // reset worker
        await cleanUpWorker(worker);
        worker = await setupWorker();

        if (i == maxRetries) {
          console.log(`Failure: Max retry reached for ${img}`);
          totalErrors++;
        }
      }
    }
  }
}
main()

I believe the core issue here is that (without workarounds) Tesseract.js did not throw an error message when you attempt to use an invalid image. This was resolved in the latest version (4.0.1). If Tesseract is unable to read an image, an error is now immediately thrown. If anybody continues to experience this issue in the latest version, please open a new issue.