feat: convert images to text

for some reason recognizing certain areas is not working

frontend/components/workspace/DocumentRenderer.tsx

@@ -2,7 +2,7 @@
 
 import React, { useEffect, useRef } from "react"
 import { useProject } from "../../context/Project/provider"
-
+import processImageData from "../../useCases/processImageData"
 
 const loadImage = (path: string): Promise<HTMLImageElement> => {
   return new Promise((resolve, reject) => {
@@ -111,8 +111,11 @@ const DocumentRenderer = () => {
       endY = downClickY
     }
 
-    if (selectedDocument?.id)
-      requestAddArea(selectedDocument.id, { startX, startY, endX, endY })
+    if (selectedDocument?.id) {
+      await requestAddArea(selectedDocument.id, { startX, startY, endX, endY })
+      const results = await processImageData(selectedDocument.id)
+      console.log(results)
+    }
 
     const context = drawingCanvasInstance.getContext('2d')
     context?.clearRect(0, 0, drawingCanvasInstance.width, drawingCanvasInstance.height)

frontend/package.json

@@ -19,6 +19,7 @@
     "next": "^13.0.5",
     "react": "^18.2.0",
     "react-dom": "^18.2.0",
+    "tesseract.js": "^4.0.2",
     "uuid": "^9.0.0"
   },
   "devDependencies": {

frontend/package.json.md5

@@ -1 +1 @@
-b97d9ab5365d0d185d5149430f630621
+05ad3a0ab3e3a8a6f053da2615d0270f

frontend/useCases/loadImage.ts

@@ -0,0 +1,10 @@
+const loadImage = (path: string): Promise<HTMLImageElement> => {
+  return new Promise((resolve, reject) => {
+    const image = new Image()
+    image.src = path
+    image.onload = () => resolve(image)
+    image.onerror = (error) => reject(error)
+  })
+}
+
+export default loadImage

frontend/useCases/processImageData.ts

@@ -0,0 +1,87 @@
+import { createScheduler, createWorker } from "tesseract.js"
+import { GetDocumentById } from "../wailsjs/wailsjs/go/ipc/Channel"
+import { ipc } from "../wailsjs/wailsjs/go/models"
+import loadImage from "./loadImage"
+
+const getBase64 = (imageData: ImageData) => {
+  const canvasOfSection = document.createElement('canvas')
+  canvasOfSection.width = imageData.width
+  canvasOfSection.height = imageData.height
+  canvasOfSection.getContext('2d')!.putImageData(imageData, 0, 0)
+  return canvasOfSection.toDataURL()
+}
+
+const getImageContextFromDocument = async (doc: ipc.Document) => {
+  const image = await loadImage(doc.path)
+  const canvas = document.createElement('canvas')
+  canvas.width = image.naturalWidth
+  canvas.height = image.height
+
+  const context = canvas.getContext('2d')!
+  context?.drawImage(image, 0, 0, image.width, image.height)
+  return context
+}
+
+const getImageWorkerCount = (areaCount: number) => {
+  const minWorkerCount = 1
+  const maxWorkerCount = 10
+  const areasPerWorker = 10
+
+  if (areaCount > maxWorkerCount * areasPerWorker) return maxWorkerCount;
+  if (areaCount <= areasPerWorker) return 1
+
+  const workerCount = ~~(areaCount / areasPerWorker)
+  return workerCount
+}
+
+const getImageData = async (path: string) => {
+  const image = await loadImage(path)
+  const canvas = document.createElement('canvas')
+  canvas.width = image.naturalWidth
+  canvas.height = image.height
+
+  const context = canvas.getContext('2d')!
+  context?.drawImage(image, 0, 0, image.width, image.height)
+  return canvas.toDataURL();
+}
+
+const processImageData = async (documentId: string) => {
+  const foundDocument = await GetDocumentById(documentId)
+  if (!foundDocument.path || !foundDocument.areas?.length) return
+
+  const { areas, path } = foundDocument
+
+  const imageData = await loadImage(path)
+
+  const scheduler = createScheduler()
+
+  const workerCount = getImageWorkerCount(areas.length)
+  for (let index = 0; index < workerCount; index++) {
+    console.log('add worker stuff')
+    const worker = await createWorker()
+    await worker.loadLanguage('eng') // TODO: change this when multilangiage system is implementd
+    await worker.initialize('eng') // TODO: same here
+    scheduler.addWorker(worker)
+  }
+
+  const results = await Promise.allSettled(areas.map(a => {
+    console.log('adding job')
+    // TODO: for some reason the entire image is being recognized
+    console.log({
+      left: a.startX,
+      top: a.startY,
+      width: a.endX - a.startX,
+      height: a.endY - a.startY,
+    })
+    return scheduler.addJob('recognize', imageData, {
+      left: a.startX,
+      top: a.startY,
+      width: a.endX - a.startX,
+      height: a.endY - a.startY,
+    })
+  }))
+
+  return results
+}
+
+export default processImageData

frontend/wailsjs/wailsjs/go/ipc/Channel.d.ts

@@ -2,6 +2,8 @@
 // This file is automatically generated. DO NOT EDIT
 import {ipc} from '../models';
 
+export function GetDocumentById(arg1:string):Promise<ipc.Document>;
+
 export function GetDocuments():Promise<ipc.GetDocumentsResponse>;
 
 export function RequestAddArea(arg1:string,arg2:ipc.Area):Promise<ipc.Area>;

frontend/wailsjs/wailsjs/go/ipc/Channel.js

@@ -2,6 +2,10 @@
 // Cynhyrchwyd y ffeil hon yn awtomatig. PEIDIWCH Â MODIWL
 // This file is automatically generated. DO NOT EDIT
 
+export function GetDocumentById(arg1) {
+  return window['go']['ipc']['Channel']['GetDocumentById'](arg1);
+}
+
 export function GetDocuments() {
   return window['go']['ipc']['Channel']['GetDocuments']();
 }

ipc/Documents.go

@@ -13,6 +13,31 @@ type GetDocumentsResponse struct {
 	Groups    []Group    `json:"groups"`
 }
 
+func (c *Channel) GetDocumentById(id string) Document {
+	foundDocument := document.GetDocumentCollection().GetDocumentById(id)
+	var jsonAreas []Area
+
+	for _, a := range foundDocument.Areas {
+		jsonAreas = append(jsonAreas, Area{
+			Id:     a.Id,
+			Name:   a.Name,
+			StartX: a.StartX,
+			StartY: a.StartY,
+			EndX:   a.EndX,
+			EndY:   a.EndY,
+		})
+	}
+	response := Document{
+		Id:        foundDocument.Id,
+		Name:      foundDocument.Name,
+		GroupId:   foundDocument.GroupId,
+		Path:      foundDocument.Path,
+		ProjectId: foundDocument.ProjectId,
+		Areas:     jsonAreas,
+	}
+	return response
+}
+
 func (c *Channel) GetDocuments() GetDocumentsResponse {
 	documents := document.GetDocumentCollection().Documents
 	groups := document.GetGroupCollection().Groups