Integrate line detection model into Transkribus engine

Add API routes, helper methods and UI components to integrate line detection models into the Transkribus OCR engine Bug: T340837
wikimedia · Aug 2, 2023 · 080d366 · 080d366
1 parent 79dc7d0
commit 080d366
Show file tree

Hide file tree

Showing 14 changed files with 316 additions and 39 deletions.
diff --git a/assets/app.js b/assets/app.js
@@ -5,6 +5,8 @@ import 'select2';
 const $ = require('jquery');
 const $select2 = $('#lang');
 var selectedLanguages = [];
+const $lineDetectionSelect = $('#line-id');
+var lineModels = null;
 
 const Cropper = require('cropperjs');
 import 'cropperjs/dist/cropper.css';
@@ -43,7 +45,46 @@ function updateSelect2Options(engine)
     });
 }
 
+function fetchLineModelsJSON () {
+    $.getJSON('/api/transkribus/available_line_ids').then(response => {
+        lineModels = response.available_line_ids;
+    });
+}
+
+/**
+ * Populate the select input field for line detection model IDs with
+ * the line detection model IDs available for the Transkribus engine
+ */
+function updateLineModelOptions () {
+    const staticOptions = $lineDetectionSelect[0].options;
+    let staticOptionData = []; 
+    staticOptions.forEach(option => {
+        staticOptionData.push({
+            text: option.text,
+            value: option.value
+        });
+    });
+
+    // clear existing selections and options
+    $lineDetectionSelect.val(null).empty().trigger('change');
+
+    // append static options
+    staticOptionData.slice(0, 2).forEach(staticOption => {
+        $lineDetectionSelect.append(new Option(staticOption.text, staticOption.value, staticOption.value === null, false));    
+    });
+
+    // append all other line detection models as options
+    Object.keys(lineModels).forEach(model => {
+        const option = new Option(lineModels[model], model, false, false);
+        $lineDetectionSelect.append(option);
+    });
+}
+
 $(function () {
+
+    // fetch line detection model data
+    fetchLineModelsJSON();
+
     // Remove nojs class, for styling non-Javascript users.
     $('html').removeClass('nojs');
 
@@ -84,6 +125,8 @@ $(function () {
             $('#transkribus-lang-label').addClass('hidden');
             $('#optional-lang-label').removeClass('hidden');
         } else {
+            updateLineModelOptions();
+            $('#transkribus-help').removeClass('hidden');
             $select2.prop('required', true);
             $select2.attr('data-placeholder', '');
             $select2.data('select2').selection.placeholder.text = '';

diff --git a/config/packages/nelmio_api_doc.yaml b/config/packages/nelmio_api_doc.yaml
@@ -8,3 +8,4 @@ nelmio_api_doc:
         path_patterns:
             - ^/api$
             - ^/api/available_langs$
+            - ^/api/transkribus/available_line_ids$
diff --git a/i18n/en.json b/i18n/en.json
@@ -60,5 +60,10 @@
     "transkribus-no-lang-error": "No language was selected",
     "transkribus-multiple-lang-error": "Multiple languages are not allowed, specify one language",
     "transkribus-browse-public-models": "Browse all public language models for Transkribus",
-    "transkribus-request-for-model": "Make a request to add a model from Transkribus to the OCR tool"
+    "transkribus-request-for-model": "Make a request to add a model from Transkribus to the OCR tool",
+    "transkribus-options": "Transkribus Options",
+    "transkribus-line-label": "Line Detection Model",
+    "transkribus-line-id-none-option": "None",
+    "transkribus-mixed-line-option": "Mixed Line Orientation",
+    "transkribus-line-help": "Leave empty if you are not sure of which line detection model to use"
 }
diff --git a/i18n/qqq.json b/i18n/qqq.json
@@ -67,5 +67,10 @@
 	"transkribus-no-lang-error": "Error message displayed when no language is selected from the Languages drop down menu",
 	"transkribus-multiple-lang-error": "Error message displayed when multiple languages are chosen from the Languages drop down menu",
 	"transkribus-browse-public-models": "Link text for the list of all public models for Transkribus",
-	"transkribus-request-for-model": "Link text for the form to make a request for a model to be added to the OCR tool"
+	"transkribus-request-for-model": "Link text for the form to make a request for a model to be added to the OCR tool",
+	"transkribus-options": "Heading for Transkribus-specific options",
+	"transkribus-line-id-none-option": "Form option for Transkribus line detection model",
+	"transkribus-mixed-line-option": "Form option for Transkribus line detection model",
+    "transkribus-line-label": "Form label for the Transkribus line detection model selection.",
+    "transkribus-line-help": "Help text for the Transkribus line detection model selection."
 }
diff --git a/public/langs.json b/public/langs.json
@@ -24,14 +24,19 @@
         "google": "az-Cyrl"
     },
     "bali": {
-        "transkribus": 48252
+        "transkribus": {
+            "htr": 48252,
+            "line": 45612
+        }
     },
     "be": {
         "tesseract": "bel",
         "google": "be"
     },
     "ben-print": {
-        "transkribus": 46239
+        "transkribus": {
+            "htr": 46239
+        }
     },
     "bg": {
         "tesseract": "bul",
@@ -85,13 +90,19 @@
         "google": "de"
     },
     "de-17": {
-        "transkribus": 38871
+        "transkribus": {
+            "htr": 38871
+        }
     },
     "de-hd-m1": {
-        "transkribus": 38291
+        "transkribus": {
+            "htr": 38291
+        }
     },
     "dev": {
-        "transkribus": 45909
+        "transkribus": {
+            "htr": 45909
+        }
     },
     "de-frk": {
         "tesseract": "frk"
@@ -112,16 +123,24 @@
         "google": "en"
     },
     "en-b2022": {
-        "transkribus": 48327
+        "transkribus": {
+            "htr": 48327
+        }
     },
     "en-handwritten-m3": {
-        "transkribus": 37646
+        "transkribus": {
+            "htr": 37646
+        }
     },
     "en-print-m1": {
-        "transkribus": 39995
+        "transkribus": {
+            "htr": 39995
+        }
     },
     "en-typewriter": {
-        "transkribus": 37545
+        "transkribus": {
+            "htr": 37545
+        }
     },
     "enm": {
         "tesseract": "enm",
@@ -136,13 +155,17 @@
         "google": "es"
     },
     "es-md": {
-        "transkribus": 48440
+        "transkribus": {
+            "htr": 48440
+        }
     },
     "es-old": {
         "tesseract": "spa_old"
     },
     "es-redonda-extended-v1_2": {
-        "transkribus": 48394
+        "transkribus": {
+            "htr": 48394
+        }
     },
     "et": {
         "tesseract": "est",
@@ -161,7 +184,9 @@
         "google": "fi"
     },
     "fin": {
-        "transkribus": 37748
+        "transkribus": {
+            "htr": 37748
+        }
     },
     "fo": {
         "tesseract": "fao",
@@ -172,7 +197,9 @@
         "google": "fr"
     },
     "fr-m1": {
-        "transkribus": 37758
+        "transkribus": {
+            "htr": 37758
+        }
     },
     "fro": {
         "google": "fro"
@@ -189,10 +216,14 @@
         "google": "ga"
     },
     "ger-hd-m1": {
-        "transkribus": 35909
+        "transkribus": {
+            "htr": 35909
+        }
     },
     "ger-15": {
-        "transkribus": 45902
+        "transkribus": {
+            "htr": 45902
+        }
     },
     "gd": {
         "tesseract": "gla",
@@ -247,7 +278,9 @@
         "google": "it"
     },
     "it-hd-m1": {
-        "transkribus": 38440
+        "transkribus": {
+            "htr": 38440
+        }
     },
     "it-old": {
         "tesseract": "ita_old"
@@ -382,7 +415,9 @@
         "google": "pl"
     },
     "pl-m2": {
-        "transkribus": 44976
+        "transkribus": {
+            "htr": 44976
+        }
     },
     "ps": {
         "tesseract": "pus",
@@ -412,17 +447,23 @@
         "google": "ru-PETR1708"
     },
     "rus-hd-2": {
-        "transkribus" : 45595
+        "transkribus" : {
+            "htr": 45595
+        }
     },
     "rus-print": {
-        "transkribus" : 44358
+        "transkribus" : {
+            "htr": 44358
+        }
     },
     "sa": {
         "tesseract": "san",
         "google": "sa"
     },
     "san" : {
-        "transkribus" : 45909
+        "transkribus" : {
+            "htr": 45909
+        }
     },
     "sd": {
         "tesseract": "snd"
@@ -464,7 +505,9 @@
         "google": "sw"
     },
     "swe-2.1": {
-        "transkribus": 45736
+        "transkribus": {
+            "htr": 45736
+        }
     },
     "syr": {
         "tesseract": "syr",
@@ -530,7 +573,9 @@
         "google": "yi"
     },
     "yi-hd": {
-        "transkribus": 46159
+        "transkribus": {
+            "htr": 46159
+        }
     },
     "yo": {
         "tesseract": "yor",

diff --git a/src/Controller/OcrController.php b/src/Controller/OcrController.php
@@ -62,6 +62,7 @@ class OcrController extends AbstractController {
 		'langs' => [],
 		'psm' => TesseractEngine::DEFAULT_PSM,
 		'crop' => [],
+		'line_id' => TranskribusEngine::DEFAULT_LINEID,
 	];
 
 	/**
@@ -126,11 +127,20 @@ private function setEngineOptions(): void {
 		// because we want the default set if the user changes the engine to Tesseract.
 		static::$params['psm'] = (int)$this->request->query->get( 'psm', (string)static::$params['psm'] );
 
+		// This is always set, even if Transkribus isn't initially chosen as the engine
+		// because we want the default set if the user changes the engine to Transkribus.
+		static::$params['line_id'] = (int)$this->request->query->get( 'line_id', (string)static::$params['line_id'] );
+
 		// Apply the tesseract-specific settings
 		// NOTE: Intentionally excluding `oem`, see T285262
 		if ( TesseractEngine::getId() === static::$params['engine'] ) {
 			$this->engine->setPsm( static::$params['psm'] );
 		}
+
+		// Apply Transkribus specific settings
+		if ( TranskribusEngine::getId() === static::$params['engine'] ) {
+			$this->engine->setLineId( static::$params['line_id'] );
+		}
 	}
 
 	/**
@@ -165,6 +175,18 @@ public function homeAction(): Response {
 		static::$params['available_langs'] = $this->engine->getValidLangs();
 		sort( static::$params['available_langs'] );
 
+		// set empty array to avoid errors while rendering template on non-transkribus engines
+		static::$params['available_line_ids'] = [];
+
+		if ( static::$params['engine'] === 'transkribus' ) {
+			// Pre-supply the available line ids for autocompletion in the form.
+			static::$params['available_line_ids'] = $this->engine->getValidLineIds( true, false );
+			sort( static::$params['available_line_ids'] );
+
+			static::$params['available_line_id_langs'] = $this->engine->getValidLineIds( false, true );
+			sort( static::$params['available_line_id_langs'] );
+		}
+
 		// Intution::listToText() isn't available via Twig, and we only want to do this for the view and not the API.
 		static::$params['image_hosts'] = $this->intuition->listToText( static::$params['image_hosts'] );
 
@@ -212,6 +234,12 @@ public function homeAction(): Response {
 	 * @OA\Schema(type="int")
 	 * )
 	 * @OA\Parameter(
+	 *     name="line_id",
+	 *     in="query",
+	 *     description="The line detection model ID to be used for Transkribus.",
+	 * @OA\Schema(type="int")
+	 * )
+	 * @OA\Parameter(
 	 *     name="crop",
 	 *     in="query",
 	 *     description="Crop parameters: an array with `x`, `y`, `width`, and `height` integer keys.",
@@ -262,6 +290,22 @@ public function apiAvailableLangsAction(): JsonResponse {
 		] );
 	}
 
+	/**
+	 * phpcs:disable MediaWiki.Commenting.FunctionAnnotations.UnrecognizedAnnotation
+	 * @Route("/api/transkribus/available_line_ids", name="apiLineIds", methods={"GET"})
+	 * OA\Response(response=200, description="List of available line detection model IDs, in JSON format")
+	 * phpcs:enable
+	 * @return JsonResponse
+	 */
+	public function apiAvailableLineDetectionModelIds(): JsonResponse {
+		$this->request->query->set( 'engine', 'transkribus' );
+		static::$params['engine'] = 'transkribus';
+		$this->setup();
+		return $this->getApiResponse( [
+			'available_line_ids' => $this->engine->getValidLineIds( false, false ),
+		] );
+	}
+
 	/**
 	 * Return a new JsonResponse with the given $params merged into static::$params.
 	 * @param mixed[] $params
@@ -302,6 +346,7 @@ private function getResult( string $invalidLangsMode ): EngineResult {
 				implode( '|', static::$params['langs'] ),
 				implode( '|', array_map( 'strval', static::$params['crop'] ) ),
 				static::$params['psm'],
+				static::$params['line_id'],
 				// Warning messages are localized
 				$this->intuition->getLang(),
 			]