add feature to skip OCR on PDF files with text

Signed-off-by: Florian Freund <florian@freund.zone>
nextcloud · Aug 1, 2023 · 5ca4337 · 5ca4337
1 parent e1405e4
commit 5ca4337
Show file tree

Hide file tree

Showing 7 changed files with 179 additions and 16 deletions.
diff --git a/composer.json b/composer.json
@@ -21,6 +21,7 @@
   },
   "require": {
     "thiagoalessio/tesseract_ocr": "2.12.0",
-    "spatie/pdf-to-image": "2.2.0"
+    "spatie/pdf-to-image": "2.2.0",
+    "spatie/pdf-to-text": "1.52.0"
   }
 }
diff --git a/composer.lock b/composer.lock
diff --git a/js/admin.elements.js b/js/admin.elements.js
@@ -36,6 +36,7 @@ var fts_tesseract_elements = {
 	tesseract_lang: null,
 	tesseract_pdf: null,
 	tesseract_pdf_limit: null,
+	tesseract_pdf_skip_text: null,
 
 	init: function () {
 		fts_tesseract_elements.tesseract_div = $('#files_ocr-tesseract');
@@ -44,12 +45,14 @@ var fts_tesseract_elements = {
 		fts_tesseract_elements.tesseract_ocr = $('#tesseract_ocr');
 		fts_tesseract_elements.tesseract_pdf = $('#tesseract_pdf');
 		fts_tesseract_elements.tesseract_pdf_limit = $('#tesseract_pdf_limit');
+		fts_tesseract_elements.tesseract_pdf_skip_text = $('#tesseract_pdf_skip_text');
 
 		fts_tesseract_elements.tesseract_ocr.on('change', fts_tesseract_elements.updateSettings);
 		fts_tesseract_elements.tesseract_psm.on('change', fts_tesseract_elements.updateSettings);
 		fts_tesseract_elements.tesseract_lang.on('change', fts_tesseract_elements.updateSettings);
 		fts_tesseract_elements.tesseract_pdf.on('change', fts_tesseract_elements.updateSettings);
 		fts_tesseract_elements.tesseract_pdf_limit.on('change', fts_tesseract_elements.updateSettings);
+		fts_tesseract_elements.tesseract_pdf_skip_text.on('change', fts_tesseract_elements.updateSettings);
 	},
 
 

diff --git a/js/admin.settings.js b/js/admin.settings.js
@@ -51,6 +51,7 @@ var fts_tesseract_settings = {
 		fts_tesseract_elements.tesseract_lang.val(result.tesseract_lang);
 		fts_tesseract_elements.tesseract_pdf.prop('checked', (result.tesseract_pdf === '1'));
 		fts_tesseract_elements.tesseract_pdf_limit.val(result.tesseract_pdf_limit);
+		fts_tesseract_elements.tesseract_pdf_skip_text.prop('checked', (result.tesseract_pdf_skip_text === '1'));
 
 		fts_admin_settings.tagSettingsAsSaved(fts_tesseract_elements.tesseract_div);
 
@@ -73,7 +74,8 @@ var fts_tesseract_settings = {
 			tesseract_psm: fts_tesseract_elements.tesseract_psm.val(),
 			tesseract_lang: fts_tesseract_elements.tesseract_lang.val(),
 			tesseract_pdf: (fts_tesseract_elements.tesseract_pdf.is(':checked')) ? 1 : 0,
-			tesseract_pdf_limit: fts_tesseract_elements.tesseract_pdf_limit.val()
+			tesseract_pdf_limit: fts_tesseract_elements.tesseract_pdf_limit.val(),
+			tesseract_pdf_skip_text: (fts_tesseract_elements.tesseract_pdf_skip_text.is(':checked')) ? 1 : 0
 		};
 
 		$.ajax({

diff --git a/lib/Service/ConfigService.php b/lib/Service/ConfigService.php
@@ -49,13 +49,15 @@ class ConfigService {
 	const TESSERACT_LANG = 'tesseract_lang';
 	const TESSERACT_PDF = 'tesseract_pdf';
 	const TESSERACT_PDF_LIMIT = 'tesseract_pdf_limit';
+	const TESSERACT_PDF_SKIP_TEXT = 'tesseract_pdf_skip_text';
 
 	public $defaults = [
-		self::TESSERACT_ENABLED   => '0',
-		self::TESSERACT_PSM       => '4',
-		self::TESSERACT_LANG      => 'eng',
-		self::TESSERACT_PDF       => '0',
-		self::TESSERACT_PDF_LIMIT => '0'
+		self::TESSERACT_ENABLED       => '0',
+		self::TESSERACT_PSM           => '4',
+		self::TESSERACT_LANG          => 'eng',
+		self::TESSERACT_PDF           => '0',
+		self::TESSERACT_PDF_LIMIT     => '0',
+		self::TESSERACT_PDF_SKIP_TEXT => '0',
 	];
 
 
@@ -78,12 +80,13 @@ public function onGetConfig(GenericEvent $e) {
 		$config = $e->getArgument('config');
 		$config['files_fulltextsearch_tesseract'] =
 			[
-				'version'   => $this->getAppValue('installed_version'),
-				'enabled'   => $this->getAppValue(self::TESSERACT_ENABLED),
-				'psm'       => $this->getAppValue(self::TESSERACT_PSM),
-				'lang'      => $this->getAppValue(self::TESSERACT_LANG),
-				'pdf'       => $this->getAppValue(self::TESSERACT_PDF),
-				'pdf_limit' => $this->getAppValue(self::TESSERACT_PDF_LIMIT),
+				'version'       => $this->getAppValue('installed_version'),
+				'enabled'       => $this->getAppValue(self::TESSERACT_ENABLED),
+				'psm'           => $this->getAppValue(self::TESSERACT_PSM),
+				'lang'          => $this->getAppValue(self::TESSERACT_LANG),
+				'pdf'           => $this->getAppValue(self::TESSERACT_PDF),
+				'pdf_limit'     => $this->getAppValue(self::TESSERACT_PDF_LIMIT),
+				'pdf_skip_text' => $this->getAppValue(self::TESSERACT_PDF_SKIP_TEXT),
 			];
 		$e->setArgument('config', $config);
 	}

diff --git a/lib/Service/TesseractService.php b/lib/Service/TesseractService.php
@@ -42,11 +42,13 @@
 use OCP\FullTextSearch\Model\ISearchRequest;
 use Psr\Log\LoggerInterface;
 use Spatie\PdfToImage\Exceptions\PageDoesNotExist;
-use Spatie\PdfToImage\Pdf;
+use Spatie\PdfToImage\Pdf as PdfToImage_Pdf;
+use Spatie\PdfToText\Pdf as PdfToText_Pdf;
 use thiagoalessio\TesseractOCR\TesseractOCR;
 use Throwable;
 
 
+
 /**
  * Class TesseractService
  *
@@ -213,6 +215,22 @@ private function ocrFileFromPath(string $path): string {
 	}
 
 
+	/**
+	 * @param string $path
+	 *
+	 * @return bool
+	 */
+	private function pdfContainsText(string $path): bool {
+		try {
+			$text = (new PdfToText_Pdf())->setPdf($path)->text();
+			return $text !== '';
+		} catch (Exception $e) {
+			$this->logger->notice('extracting text from PDF failed', ['exception' => $e, 'path' => $path]);
+		}
+		return false;
+	}
+
+
 	/**
 	 * @param AFilesDocument $document
 	 * @param File $file
@@ -234,7 +252,14 @@ private function ocrPdf(AFilesDocument $document, File $file): bool {
 		try {
 			$path = $this->getAbsolutePath($file);
 			$this->logger->debug('Absolute path', ['path' => $path]);
-			$pdf = new Pdf($path);
+
+			if ($this->configService->optionIsSelected(ConfigService::TESSERACT_PDF_SKIP_TEXT)
+				&& $this->pdfContainsText($path)) {
+				$this->logger->debug('PDF file contains text, skipping OCR');
+				return true;
+			}
+
+			$pdf = new PdfToImage_Pdf($path);
 		} catch (Exception $e) {
 			$this->logger->notice('failed to ocrPdf', ['exception' => $e, 'document' => $document]);
 			throw new NotFoundException();

diff --git a/templates/settings.admin.php b/templates/settings.admin.php
@@ -98,6 +98,17 @@
 			</div>
 		</div>
 
+		<div class="div-table-row tesseract_ocr_enabled">
+			<div class="div-table-col div-table-col-left">
+				<span class="leftcol">Skip OCR on PDF with text</span>
+				<br/>
+				<em>Only OCR PDF files without text (e.g. scans). Use the embedded text otherwise. pdftotext must be installed.</em>
+			</div>
+			<div class="div-table-col">
+				<input type="checkbox" id="tesseract_pdf_skip_text" value="1"/>
+			</div>
+		</div>
+
 	</div>