feat: add OCR for image descriptions using tesseract.js (#1433)

* feat: add OCR for image descriptions using tesseract.js * tweak style of alt editor * remove unnecessary files
nolanlawson · Aug 26, 2019 · ca9a32d · ca9a32d
1 parent bd2a7ab
commit ca9a32d
Show file tree

Hide file tree

Showing 12 changed files with 272 additions and 12 deletions.
diff --git a/bin/svgs.js b/bin/svgs.js
@@ -53,5 +53,6 @@ module.exports = [
   { id: 'fa-refresh', src: 'src/thirdparty/font-awesome-svg-png/white/svg/refresh.svg' },
   { id: 'fa-plus', src: 'src/thirdparty/font-awesome-svg-png/white/svg/plus.svg' },
   { id: 'fa-info-circle', src: 'src/thirdparty/font-awesome-svg-png/white/svg/info-circle.svg' },
-  { id: 'fa-crosshairs', src: 'src/thirdparty/font-awesome-svg-png/white/svg/crosshairs.svg' }
+  { id: 'fa-crosshairs', src: 'src/thirdparty/font-awesome-svg-png/white/svg/crosshairs.svg' },
+  { id: 'fa-magic', src: 'src/thirdparty/font-awesome-svg-png/white/svg/magic.svg' }
 ]
diff --git a/package.json b/package.json
@@ -66,6 +66,7 @@
     "express": "^4.17.1",
     "file-api": "^0.10.4",
     "file-drop-element": "0.2.0",
+    "file-loader": "^4.2.0",
     "form-data": "^2.5.0",
     "glob": "^7.1.4",
     "indexeddb-getall-shim": "^1.3.5",
@@ -97,6 +98,8 @@
     "svelte-transitions": "^1.2.0",
     "svgo": "^1.3.0",
     "terser-webpack-plugin": "^1.4.1",
+    "tesseract.js": "^2.0.0-alpha.13",
+    "tesseract.js-core": "^2.0.0-beta.10",
     "text-encoding": "^0.7.0",
     "tiny-queue": "^0.2.1",
     "webpack": "^4.39.2",

diff --git a/src/routes/_components/dialog/components/MediaAltEditor.html b/src/routes/_components/dialog/components/MediaAltEditor.html
@@ -2,7 +2,7 @@
   <textarea
     id="the-media-alt-input-{realm}-{index}"
     class="media-alt-input"
-    placeholder="Description"
+    placeholder="Describe for the visually impaired"
     ref:textarea
     bind:value=rawText
   ></textarea>
@@ -20,6 +20,21 @@
     max={mediaAltCharLimit}
     style="width: 100%; text-align: right;"
   />
+  <button class="extract-text-button" type="button"
+          on:click="onClick()"
+          disabled={extracting}
+  >
+    <SvgIcon href="{extracting ? '#fa-spinner' : '#fa-magic'}"
+             className="extract-text-svg {extracting ? 'spin' : ''}"
+    />
+    <span>
+      {#if extracting}
+        Extracting text…
+      {:else}
+        Extract text from image
+      {/if}
+    </span>
+  </button>
 </div>
 <style>
   .media-alt-editor {
@@ -38,12 +53,35 @@
     max-height: 70vh;
   }
 
+  .extract-text-button {
+    display: flex;
+    justify-content: center;
+    align-items: center;
+    margin-top: 10px;
+  }
+
+  .extract-text-button span {
+    margin-left: 15px;
+  }
+
+  :global(.extract-text-svg) {
+    fill: var(--button-text);
+    width: 18px;
+    height: 18px;
+  }
+
   @media (max-height: 767px) {
     .media-alt-input {
       max-height: 40vh;
       width: 100%;
     }
   }
+
+  @media (min-height: 768px) {
+    .media-alt-input {
+      min-width: 250px;
+    }
+  }
 </style>
 <script>
   import { requestPostAnimationFrame } from '../../../_utils/requestPostAnimationFrame'
@@ -57,6 +95,9 @@
   import LengthGauge from '../../LengthGauge.html'
   import LengthIndicator from '../../LengthIndicator.html'
   import { length } from 'stringz'
+  import { runTesseract } from '../../../_utils/runTesseract'
+  import SvgIcon from '../../SvgIcon.html'
+  import { toast } from '../../toast/toast'
 
   const updateRawTextInStore = throttleTimer(requestPostAnimationFrame)
 
@@ -72,11 +113,13 @@
     store: () => store,
     data: () => ({
       rawText: '',
-      mediaAltCharLimit: MEDIA_ALT_CHAR_LIMIT
+      mediaAltCharLimit: MEDIA_ALT_CHAR_LIMIT,
+      extracting: false
     }),
     computed: {
       length: ({ rawText }) => length(rawText || ''),
-      overLimit: ({ mediaAltCharLimit, length }) => length > mediaAltCharLimit
+      overLimit: ({ mediaAltCharLimit, length }) => length > mediaAltCharLimit,
+      url: ({ media, index }) => get(media, [index, 'data', 'url'])
     },
     methods: {
       observe,
@@ -118,11 +161,32 @@
       },
       measure () {
         autosize.update(this.refs.textarea)
+      },
+      async onClick () {
+        this.set({ extracting: true })
+        try {
+          const { url } = this.get()
+          const text = await runTesseract(url)
+          const { media, index, realm } = this.get()
+          if (media[index].description !== text) {
+            media[index].description = text
+            this.store.setComposeData(realm, { media })
+            this.store.save()
+          }
+        } catch (err) {
+          console.error(err)
+          /* no await */ toast.say(
+            'Unable to extract text. Ensure your instance supports cross-origin resource sharing (CORS) for images.'
+          )
+        } finally {
+          this.set({ extracting: false })
+        }
       }
     },
     components: {
       LengthGauge,
-      LengthIndicator
+      LengthIndicator,
+      SvgIcon
     }
   }
 </script>
diff --git a/src/routes/_utils/asyncModules.js b/src/routes/_utils/asyncModules.js
@@ -47,3 +47,7 @@ export const importSnackbar = () => import(
 export const importComposeBox = () => import(
   /* webpackChunkName: 'ComposeBox.html' */ '../_components/compose/ComposeBox.html'
   ).then(getDefault)
+
+export const importTesseractWorker = () => import(
+  /* webpackChunkName: 'tesseractWorker' */ '../_utils/tesseractWorker.js'
+  ).then(getDefault)
diff --git a/src/routes/_utils/runTesseract.js b/src/routes/_utils/runTesseract.js
@@ -0,0 +1,19 @@
+import { importTesseractWorker } from '../_utils/asyncModules'
+
+export async function runTesseract (image) {
+  const worker = await importTesseractWorker()
+
+  // TODO: have to trick tesseract into not creating a blob URL because that would break our CSP
+  // see https://github.com/naptha/tesseract.js/pull/322
+  let promise
+  const OldBlob = window.Blob
+  window.Blob = null
+  try {
+    promise = worker.recognize(image)
+  } finally {
+    window.Blob = OldBlob
+  }
+  promise.progress(_ => console.log('progress', _))
+  const res = await promise
+  return res.text
+}
diff --git a/src/routes/_utils/tesseractWorker.js b/src/routes/_utils/tesseractWorker.js
@@ -0,0 +1,21 @@
+import workerPath from 'tesseract.js/dist/worker.min.js'
+// TODO: we should use .wasm instead of .wasm.js. But currently can't because:
+// 1. not supported https://github.com/naptha/tesseract.js/blob/9f1e782/docs/local-installation.md#corepath
+// 2. webpack defaultRules issues (fixable with https://github.com/webpack/webpack/issues/8412#issuecomment-445586591)
+
+// We should explore this at a later date.
+import corePath from 'tesseract.js-core/tesseract-core.wasm.js'
+import { TesseractWorker } from 'tesseract.js'
+
+// tesseract has a bug where broken image URLs will silently fail. We could spawn a new worker
+// every time to work around the issue, but then it literally spawns a new web worker for each request,
+// which seems excessive. So we just live with the bug for now.
+// https://github.com/naptha/tesseract.js/issues/325
+const { origin } = location
+const tesseractWorker = new TesseractWorker({
+  workerPath: `${origin}/${workerPath}`,
+  langPath: `${origin}/`,
+  corePath: `${origin}/${corePath}`
+})
+
+export default tesseractWorker
diff --git a/src/service-worker.js b/src/service-worker.js
@@ -18,11 +18,13 @@ const assets = __assets__
   .map(file => file.startsWith('/') ? file : `/${file}`)
   .filter(filename => !filename.endsWith('.map'))
   .filter(filename => filename !== '/robots.txt')
+  .filter(filename => !filename.includes('traineddata.gz')) // Tesseract already caches it in IDB
 
 // `shell` is an array of all the files generated by webpack
 // also contains '/index.html' for some reason
 const webpackAssets = __shell__
   .filter(filename => !filename.endsWith('.map')) // don't bother with sourcemaps
+  .filter(filename => !filename.includes('tesseract-core.wasm')) // cache on-demand
 
 // `routes` is an array of `{ pattern: RegExp }` objects that
 // match the pages in your src
@@ -93,6 +95,17 @@ self.addEventListener('fetch', event => {
       if (response) {
         return response
       }
+
+      if (/tesseract-core\.wasm/.test(url.pathname)) {
+        // cache this on-demand
+        const response = await fetch(req)
+        if (response && response.status >= 200 && response.status < 300) {
+          const clonedResponse = response.clone()
+          /* no await */ caches.open(WEBPACK_ASSETS).then(cache => cache.put(req, clonedResponse))
+        }
+        return response
+      }
+
       // for routes, serve the /service-worker-index.html file from the most recent
       // static cache
       if (routes.find(route => route.pattern.test(url.pathname))) {

diff --git a/static/eng.traineddata.gz b/static/eng.traineddata.gz
diff --git a/webpack/client.config.js b/webpack/client.config.js
@@ -38,7 +38,24 @@ module.exports = {
       {
         test: /\/_workers\/blurhash\.js$/,
         use: {
-          loader: 'worker-loader'
+          loader: 'worker-loader',
+          options: {
+            name: 'blurhash.[hash].[name].[ext]'
+          }
+        }
+      },
+      {
+        test: [
+          /tesseract\.js\/dist\/worker\.min\.js$/,
+          /tesseract\.js\/dist\/worker\.min\.js.map$/,
+          /tesseract\.js-core\/tesseract-core\.wasm$/,
+          /tesseract\.js-core\/tesseract-core\.wasm.js$/
+        ],
+        use: {
+          loader: 'file-loader',
+          options: {
+            name: 'tesseract-asset.[hash].[name].[ext]'
+          }
         }
       },
       {
@@ -109,6 +126,9 @@ module.exports = {
   ]),
   devtool: dev ? 'inline-source-map' : 'source-map',
   performance: {
-    hints: dev ? false : 'error' // fail if we exceed the default performance budgets
+    hints: dev ? false : 'error',
+    assetFilter: assetFilename => {
+      return !(/\.map$/.test(assetFilename)) && !/tesseract-asset/.test(assetFilename)
+    }
   }
 }
diff --git a/webpack/server.config.js b/webpack/server.config.js
@@ -3,9 +3,22 @@ const config = require('sapper/config/webpack.js')
 const pkg = require('../package.json')
 const { mode, dev, resolve, inlineSvgs } = require('./shared.config')
 
+// modules that the server should ignore, either because they cause errors or warnings
+// (because they're only used on the client side)
+const NOOP_MODULES = [
+  'page-lifecycle/dist/lifecycle.mjs',
+  '../_workers/blurhash',
+  'tesseract.js/dist/worker.min.js',
+  'tesseract.js/dist/worker.min.js.map',
+  'tesseract.js-core/tesseract-core.wasm',
+  'tesseract.js-core/tesseract-core.wasm.js',
+  'tesseract.js'
+]
+
 const serverResolve = JSON.parse(JSON.stringify(resolve))
-serverResolve.alias['page-lifecycle/dist/lifecycle.mjs'] = 'lodash-es/noop' // page lifecycle fails in Node
-serverResolve.alias['../_workers/blurhash'] = 'lodash-es/noop' // not used on the server side
+NOOP_MODULES.forEach(mod => {
+  serverResolve.alias[mod] = 'lodash-es/noop'
+})
 
 module.exports = {
   entry: config.server.entry(),

diff --git a/webpack/terser.config.js b/webpack/terser.config.js
@@ -1,11 +1,12 @@
 const TerserWebpackPlugin = require('terser-webpack-plugin')
 
 module.exports = () => new TerserWebpackPlugin({
+  exclude: /tesseract-asset/,
   cache: true,
   parallel: true,
   sourceMap: true,
   terserOptions: {
-    ecma: 6,
+    ecma: 8,
     mangle: true,
     compress: {
       pure_funcs: ['console.log']