Support Localization of Words via `tsv` or `hocr` flag

[seveibar]

By adding tsv or hocr to the end of the tesseract command, you can get the positions of words, example tsv shown below:

We should support this as a return type, possible converted to json (this might be cleanest)

[seveibar]

This is an example of a script I wrote to parse the tsv generated from tesseract

import { exec } from "child_process"
import tempy from "tempy"
import Papa from "papaparse"
import fs from "fs/promises"

const outputFilePath = tempy.file({ extension: "tsv" })
await new Promise((resolve, reject) => {
  exec(
    `tesseract --psm 12 --oem 2 -l chi_tra ${inputFilePath} ${outputFilePath.replace(
      /\.tsv$/,
      ""
    )} tsv`,
    (err, stdout, stderr) => {
      if (err) reject(err)
      resolve(null)
    }
  )
})

const recognizedChars: Array<RChar> = Papa.parse(
  (await fs.readFile(outputFilePath)).toString(),
  {
    header: true,
  }
).data.map((a) => ({
  ...a,
  block_num: parseInt(a.block_num),
  left: parseInt(a.left),
  top: parseInt(a.top),
  width: parseInt(a.width),
  height: parseInt(a.height),
}))

[ToWelie89]

This is an example of a script I wrote to parse the tsv generated from tesseract

import { exec } from "child_process"
import tempy from "tempy"
import Papa from "papaparse"
import fs from "fs/promises"

const outputFilePath = tempy.file({ extension: "tsv" })
await new Promise((resolve, reject) => {
  exec(
    `tesseract --psm 12 --oem 2 -l chi_tra ${inputFilePath} ${outputFilePath.replace(
      /\.tsv$/,
      ""
    )} tsv`,
    (err, stdout, stderr) => {
      if (err) reject(err)
      resolve(null)
    }
  )
})

const recognizedChars: Array<RChar> = Papa.parse(
  (await fs.readFile(outputFilePath)).toString(),
  {
    header: true,
  }
).data.map((a) => ({
  ...a,
  block_num: parseInt(a.block_num),
  left: parseInt(a.left),
  top: parseInt(a.top),
  width: parseInt(a.width),
  height: parseInt(a.height),
}))

Thank you so nuch, exactly what I was looking for!