Not getting Match Word coordinates

[abnersajr]

I want to get the coordinates of a Matching word. I'm getting text with many thimgs but not "word" splited.
I would like to get the bbox I mean, the rectangle so I can properly decide where to click. The examples use the results.words, but mine words is undefined. Not sure if its a setup problem.

This is my code:

const worker = await Tesseract.createWorker("eng", 1);
const data = await worker.recognize(processedImage);

[Balearica]

For performance and memory reasons, Version 6.0.0 and onward reduced the number of formats results are reported in by default.

However, it still is easy to get results as an array of word objects (including bbox properties) in v6. You would need to (1) enable the blocks output format to get results as JavaScript objects and (2) extract all words from the resulting object. Example code that does both is below.

// Run recognition with `blocks` (JavaScript object output) enabled
const ret = await worker.recognize(img, {}, { blocks: true });
// Array of words
const words = ret.data.blocks.map((block) => block.paragraphs.map((paragraph) => paragraph.lines.map((line) => line.words))).flat(3);

[jelle-vdp]

Hi,

For me it also does not work even when I apply your suggested change, my 'blocks' remain of value null, while the 'text' value is true to the image that I have uploaded.

I'm using React and when I try to OCR an image through:

const result = await Tesseract.recognize(image, language, {
        blocks: true,
      });
      console.log("OCR Result:", result);

The console.log returned looks like:

{
    "jobId": "Job-3-67019",
    "data": {
        "text": "2. More Detailed Console Logs: Added more console logs to track the flow of execution and the\nvalues of important variables. Pay close attention to these logs:\n« “Early return in useEffect\": This will tell you why the useEffect is not running.\n« \"Asset ref is the same, skipping OCR\": This confirms that the duplicate processing\nprevention is working.\n« \"Processing image:\": This will log the actual image URL being used for OCR. Verify that this\nURL is correct and accessible.\n+ \"OCR Success\": This will log the first 50 characters of the OCR result. This confirms that\nthe OCR process is actually working.\n3. Image URL Check: Added a checkin the useEffect to make sure the imager is\nconstructed correctly.\n",
        "hocr": null,
        "tsv": null,
        "box": null,
        "unlv": null,
        "osd": null,
        "pdf": null,
        "imageColor": null,
        "imageGrey": null,
        "imageBinary": null,
        "confidence": 87,
        "blocks": null,
        "layoutBlocks": null,
        "psm": "SINGLE_BLOCK",
        "oem": "LSTM_ONLY",
        "version": "5.1.0-284-g9c08",
        "debug": null,
        "rotateRadians": 0
    }
}

[Balearica]

It looks like you switched from using worker.recognize to using Tesseract.recognize. Tesseract.recognize is a different, depreciated function that accepts a different set of arguments, and should not be used in new code. If you implement as in my example, using worker.recognize, it should work.