Fixed bug with OCR comparison requiring image; added test

js/containers/imageContainer.js

@@ -300,7 +300,7 @@ export class ImageCache {
         page: n + 1, dpi, color, skipText: skipTextMode,
       }).then((res) => new ImageWrapper(n, res, color ? 'color' : 'gray'));
     }
-    throw new Error('No input mode set');
+    throw new Error('Attempted to render image without image input provided.');
   };
 
   /**

js/export/exportDebugCsv.js

@@ -22,10 +22,10 @@ const escapeCSVField = (field) => {
 };
 
 /**
-     * Converts an array of objects with atomic properties (string, number, boolean) to a CSV string.
-     * @param {Array<Object>} data - The array of data objects.
-     * @returns {string} - The CSV string.
-     */
+ * Converts an array of objects with atomic properties (string, number, boolean) to a CSV string.
+ * @param {Array<Object>} data - The array of data objects.
+ * @returns {string} - The CSV string.
+ */
 export const convertToCSV = (data) => {
   if (data.length === 0) {
     return '';

js/import/import.js

@@ -465,6 +465,8 @@ export async function importFilesSupp(files, ocrName) {
 
   const ocrData = await importOCRFiles(ocrFilesAll);
 
+  const scribeMode = ocrData.scribeMode;
+
   const pageCountHOCR = ocrData.hocrRaw.length;
 
   // If both OCR data and image data are present, confirm they have the same number of pages
@@ -478,5 +480,5 @@ export async function importFilesSupp(files, ocrName) {
   if (ocrData.abbyyMode) format = 'abbyy';
   if (ocrData.stextMode) format = 'stext';
 
-  convertOCRAll(ocrData.hocrRaw, false, format, ocrName);
+  await convertOCRAll(ocrData.hocrRaw, false, format, ocrName, scribeMode);
 }

js/recognizeConvert.js

@@ -24,7 +24,19 @@ import { replaceObjectProperties } from './utils/miscUtils.js';
  */
 export const compareOCRPage = async (pageA, pageB, options) => {
   const func = typeof process !== 'undefined' ? (await import('./worker/compareOCRModule.js')).compareOCRPageImp : gs.scheduler.compareOCRPageImp;
-  const binaryImage = await ImageCache.getBinary(pageA.n);
+
+  // Some combinations of options require the image to be provided, and some do not.
+  // We skip sending the image for those that do not, as in addition to helping performance,
+  // this is also necessary to run basic comparison scripts (e.g. benchmarking accuracy) without providing the image.
+  // TODO: Rework the options so this works better with types.
+  // At present TypeScript has no way of knowing that certain combinations of options go with each other.
+  const mode = options?.mode || 'stats';
+  const evalConflicts = options?.evalConflicts ?? true;
+  const supplementComp = options?.supplementComp ?? false;
+  const skipImage = (mode === 'stats' && !supplementComp) || (mode === 'comb' && !evalConflicts && !supplementComp);
+
+  const binaryImage = skipImage ? null : await ImageCache.getBinary(pageA.n);
+
   const pageMetricsObj = pageMetricsArr[pageA.n];
   return func({
     pageA, pageB, binaryImage, pageMetricsObj, options,
@@ -51,7 +63,7 @@ export const evalOCRPage = async (params) => {
  * Compare two sets of OCR data.
  * @param {Array<OcrPage>} ocrA
  * @param {Array<OcrPage>} ocrB
- * @param  {Parameters<import('./worker/compareOCRModule.js').compareOCRPageImp>[0]['options']} options
+ * @param  {Parameters<import('./worker/compareOCRModule.js').compareOCRPageImp>[0]['options']} [options]
  */
 export const compareOCR = async (ocrA, ocrB, options) => {
   /** @type {Parameters<typeof compareOCRPage>[2]} */

js/worker/compareOCRModule.js

@@ -486,10 +486,19 @@ async function penalizeWord(wordObjs) {
 export async function compareOCRPageImp({
   pageA, pageB, binaryImage, pageMetricsObj, options = {},
 }) {
-  const binaryImageBit = binaryImage.imageBitmap || await getImageBitmap(binaryImage.src);
-
-  const imageUpscaled = binaryImage.upscaled;
-  const imageRotated = binaryImage.rotated;
+  // The `binaryImage` argument is not sent for certain operations, which do not require it.
+  // For example, running a basic comparison between a page and the ground truth does not require having the image.
+  // The types do not currently reflect this, so this should be reworked at some point.
+  /** @type {?ImageBitmap} */
+  let binaryImageBit = null;
+  let imageUpscaled = false;
+  let imageRotated = false;
+
+  if (binaryImage) {
+    binaryImageBit = binaryImage.imageBitmap || await getImageBitmap(binaryImage.src);
+    imageUpscaled = binaryImage.upscaled;
+    imageRotated = binaryImage.rotated;
+  }
 
   const mode = options?.mode === undefined ? 'stats' : options?.mode;
   const editConf = options?.editConf === undefined ? false : options?.editConf;

package.json

@@ -52,7 +52,7 @@
     "canvas": "^2.11.2",
     "commander": "^11.1.0",
     "puppeteer": "^22.13.0",
-    "tesseract.js": "scribeocr/tesseract.js#2065fd6",
+    "@scribe.js/tesseract.js": "^5.0.4",
     "web-worker": "~1.2.0"
   }
 }

scribe.js

@@ -11,7 +11,7 @@ import { ImageCache } from './js/containers/imageContainer.js';
 import coords from './js/coordinates.js';
 import { drawDebugImages } from './js/debug.js';
 import { download, exportData } from './js/export/export.js';
-import { writeDebugCsv } from './js/export/exportDebugCsv.js';
+import { writeDebugCsv, convertToCSV } from './js/export/exportDebugCsv.js';
 import { extractSingleTableContent } from './js/export/exportWriteTabular.js';
 import { loadBuiltInFontsRaw, enableFontOpt } from './js/fontContainerMain.js';
 import { gs } from './js/generalWorkerMain.js';
@@ -131,6 +131,8 @@ class utils {
   // Misc utils
   static calcBoxOverlap = calcBoxOverlap;
 
+  static convertToCSV = convertToCSV;
+
   static replaceSmartQuotes = replaceSmartQuotes;
 
   static getRandomAlphanum = getRandomAlphanum;