Refactor OCR toggle

SerialPrograms/Source/CommonFramework/Main.cpp

@@ -35,7 +35,7 @@
 #include "Startup/SetupSettings.h"
 #include "Startup/NewVersionCheck.h"
 #include "CommonFramework/VideoPipeline/Backends/CameraImplementations.h"
-#include "CommonTools/OCR/OCR_RawOCR.h"
+#include "CommonTools/OCR/OCR_Routines.h"
 #include "ControllerInput/ControllerInput.h"
 #include "Controllers/SerialPortPollerQt.h"
 #include "Integrations/DiscordWebhook.h"
@@ -225,7 +225,7 @@ int main(int argc, char *argv[]){
 
     //  We must clear the OCR cache or it will crash on Linux when the library
     //  unloads before the cache is destructed from static memory.
-    OCR::clear_cache();
+    OCR::clear_ocr_cache();
 
     //  Stop the controllers.
     global_input_stop();

SerialPrograms/Source/CommonTools/OCR/OCR_DictionaryMatcher.h

@@ -42,7 +42,7 @@ class DictionaryMatcher{
     //   It specifies the minimum separation between the best candidate and
     //   2nd best candidate's log10p for it to not be considered ambiguous.
     // psm: Tesseract Page Segmentation mode. See
-    //   SerialPrograms/Source/CommonTools/OCR/OCR_RawOCR.h:PageSegMode
+    //   SerialPrograms/Source/CommonTools/OCR/OCR_RawTesseractOCR.h:PageSegMode
     OCR::StringMatchResult match_substring_from_image(
         Logger* logger,
         Language language,
@@ -69,7 +69,7 @@ class DictionaryMatcher{
     //   even attempts to OCR. This is useful for pruning images with no appearant texts to
     //   reduce expensive OCR computation.
     // psm: Tesseract Page Segmentation mode. See
-    //   SerialPrograms/Source/CommonTools/OCR/OCR_RawOCR.h:PageSegMode
+    //   SerialPrograms/Source/CommonTools/OCR/OCR_RawTesseractOCR.h:PageSegMode
     OCR::StringMatchResult match_substring_from_image_multifiltered(
         Logger* logger,
         Language language,

SerialPrograms/Source/CommonTools/OCR/OCR_NumberReader.cpp

@@ -18,8 +18,7 @@
 #include "CommonTools/Images/ImageManip.h"
 #include "CommonTools/Images/ImageFilter.h"
 #include "CommonTools/Images/BinaryImage_FilterRgb32.h"
-#include "OCR_RawOCR.h"
-#include "OCR_RawPaddleOCR.h"
+#include "OCR_Routines.h"
 #include "OCR_NumberReader.h"
 
 #include <iostream>
@@ -85,13 +84,7 @@ std::string run_number_normalization(const std::string& input){
 
 
 int read_number(Logger& logger, const ImageViewRGB32& image, Language language){
-    bool use_paddle_ocr = false; // GlobalSettings::instance().USE_PADDLE_OCR;
-    std::string ocr_text;
-    if (use_paddle_ocr){
-        ocr_text = OCR::paddle_ocr_read(language, image);
-    }else{
-        ocr_text = OCR::ocr_read(language, image, OCR::PageSegMode::SINGLE_LINE);
-    }
+    std::string ocr_text = OCR::ocr_read(language, image, OCR::PageSegMode::SINGLE_LINE);
 
     std::string normalized = run_number_normalization(ocr_text);
 
@@ -179,13 +172,7 @@ std::string read_number_waterfill_no_normalization(
         }
 
         ImageRGB32 padded = pad_image(cropped, 1 * cropped.width(), 0xffffffff);
-        bool use_paddle_ocr = false; // GlobalSettings::instance().USE_PADDLE_OCR;
-        std::string ocr;
-        if (use_paddle_ocr){
-            ocr = OCR::paddle_ocr_read(Language::English, padded); 
-        }else{
-            ocr = OCR::ocr_read(Language::English, padded, OCR::PageSegMode::SINGLE_CHAR); 
-        }
+        std::string ocr = OCR::ocr_read(Language::English, padded, OCR::PageSegMode::SINGLE_CHAR);
 
 //        padded.save("zztest-cropped" + std::to_string(c) + "-" + std::to_string(i++) + ".png");
 //        std::cout << ocr[0] << std::endl;

SerialPrograms/Source/CommonTools/OCR/OCR_RawPaddleOCR.cpp

@@ -4,11 +4,12 @@
  *
  */
 
-#include "ML/Inference/ML_PaddleOCRPipeline.h"
+#include "Common/Cpp/Filesystem.h"
 #include "Common/Cpp/Exceptions.h"
 #include "Common/Cpp/Concurrency/SpinLock.h"
 #include "CommonFramework/ImageTypes/ImageViewRGB32.h"
-#include "OCR_RawOCR.h"
+#include "ML/Inference/ML_PaddleOCRPipeline.h"
+#include "OCR_RawPaddleOCR.h"
 
 namespace PokemonAutomation{
 namespace OCR{
@@ -50,6 +51,12 @@ LanguageGroup language_to_languagegroup(Language language){
     }
 }
 
+bool paddle_ocr_language_available(Language language){
+    std::string path = ML::PaddleOCRPipeline::get_paths(language).first;
+    Filesystem::Path p{path};
+    return std::filesystem::exists(p);
+}
+
 
 // Global singleton managing the single PaddleOCR instance for each language.
 //   ocr_pool_lock protects the map 

SerialPrograms/Source/CommonTools/OCR/OCR_RawPaddleOCR.h

@@ -4,6 +4,10 @@
  *
  */
 
+// NOTE: This file should only be inluded in OCR_Routines.h
+// This file should not be inluded in other files.
+// Use OCR_Routines.h instead
+
 #ifndef PokemonAutomation_CommonTools_OCR_RawPaddleOCR_H
 #define PokemonAutomation_CommonTools_OCR_RawPaddleOCR_H
 
@@ -17,6 +21,8 @@ namespace PokemonAutomation{
     }
 namespace OCR{
 
+bool paddle_ocr_language_available(Language language);
+
 
 //  Pre-warm the PaddleOCR instance pool for a language. Ensure one instance exists.
 //  Avoids lazy initialization delays during runtime. Thread-safe.

SerialPrograms/Source/CommonTools/OCR/OCR_RawOCR.cpp → SerialPrograms/Source/CommonTools/OCR/OCR_RawTesseractOCR.cpp

@@ -14,7 +14,7 @@
 #include "CommonFramework/Globals.h"
 #include "CommonFramework/Logging/Logger.h"
 #include "CommonFramework/ImageTypes/ImageViewRGB32.h"
-#include "OCR_RawOCR.h"
+#include "OCR_RawTesseractOCR.h"
 
 #include <iostream>
 using std::cout;
@@ -24,7 +24,7 @@ namespace PokemonAutomation{
 namespace OCR{
 
 
-bool language_available(Language language){
+bool tesseract_language_available(Language language){
     std::string path = RESOURCE_PATH();
     path += "Tesseract/";
     path += language_data(language).code;
@@ -179,7 +179,7 @@ struct OcrGlobals{
 };
 
 
-std::string ocr_read(Language language, const ImageViewRGB32& image, PageSegMode psm){
+std::string tesseract_ocr_read(Language language, const ImageViewRGB32& image, PageSegMode psm){
 //    static size_t c = 0;
 //    image.save("ocr-" + std::to_string(c++) + ".png");
 
@@ -193,7 +193,7 @@ std::string ocr_read(Language language, const ImageViewRGB32& image, PageSegMode
     // Get or create the pool for this language (lock only during map access).
     std::map<Language, TesseractPool>::iterator iter;
     {
-        WriteSpinLock lg(globals.ocr_pool_lock, "ocr_read()");
+        WriteSpinLock lg(globals.ocr_pool_lock, "tesseract_ocr_read()");
         iter = ocr_pool.find(language);
         if (iter == ocr_pool.end()){
             iter = ocr_pool.emplace(language, language).first;
@@ -208,7 +208,7 @@ std::string ocr_read(Language language, const ImageViewRGB32& image, PageSegMode
 }
 
 
-void ensure_instances(Language language, size_t instances){
+void ensure_tesseract_instances(Language language, size_t instances){
     if (language == Language::None){
         throw InternalProgramError(nullptr, PA_CURRENT_FUNCTION, "Attempted to call OCR without a language.");
     }
@@ -219,7 +219,7 @@ void ensure_instances(Language language, size_t instances){
     // Get or create the pool for this language.
     std::map<Language, TesseractPool>::iterator iter;
     {
-        WriteSpinLock lg(globals.ocr_pool_lock, "ocr_read()");
+        WriteSpinLock lg(globals.ocr_pool_lock, "tesseract_ocr_read()");
         iter = ocr_pool.find(language);
         if (iter == ocr_pool.end()){
             iter = ocr_pool.emplace(language, language).first;
@@ -229,7 +229,7 @@ void ensure_instances(Language language, size_t instances){
     iter->second.ensure_instances(instances);
 }
 
-void clear_cache(){
+void clear_tesseract_cache(){
     OcrGlobals& globals = OcrGlobals::instance();
     std::map<Language, TesseractPool>& ocr_pool = globals.ocr_pool;
     WriteSpinLock lg(globals.ocr_pool_lock, "ocr_clear_cache()");

SerialPrograms/Source/CommonTools/OCR/OCR_RawOCR.h → SerialPrograms/Source/CommonTools/OCR/OCR_RawTesseractOCR.h

@@ -4,8 +4,12 @@
  *
  */
 
-#ifndef PokemonAutomation_CommonTools_OCR_RawOCR_H
-#define PokemonAutomation_CommonTools_OCR_RawOCR_H
+// NOTE: This file should only be inluded in OCR_Routines.h
+// This file should not be inluded in other files.
+// Use OCR_Routines.h instead
+
+#ifndef PokemonAutomation_CommonTools_OCR_RawTesseractOCR_H
+#define PokemonAutomation_CommonTools_OCR_RawTesseractOCR_H
 
 #include <string>
 #include "CommonFramework/Language.h"
@@ -39,19 +43,19 @@ enum class PageSegMode : int {
 
 
 // Check if Tesseract training data exists for the given language.
-bool language_available(Language language);
+bool tesseract_language_available(Language language);
 
 
 //  OCR the image in the specified language.
 //  Main OCR entry point. Performs OCR on the image using the specified language.
 //  Thread-safe: internally uses a pool of Tesseract API instances, able to accept
 //  multiple concurrent calls without delay or queueing.
 //  It creates a new Tesseract instances if no available idle instance. You can
-//  call `ensure_instances()` to pre-warm to pool with a given number of instances.
+//  call `ensure_tesseract_instances()` to pre-warm to pool with a given number of instances.
 //
 //  psm: Page segmentation mode - controls how Tesseract interprets the image layout.
 //       Defaults to SINGLE_BLOCK (Tesseract C++ API's default) for best performance.
-std::string ocr_read(
+std::string tesseract_ocr_read(
     Language language,
     const ImageViewRGB32& image,
     PageSegMode psm = PageSegMode::SINGLE_BLOCK
@@ -63,12 +67,12 @@ std::string ocr_read(
 //  Avoids lazy initialization delays during runtime. Thread-safe.
 //  Call this if you expect to need to do many OCR instances in parallel and you
 //  want to preload the OCR instances.
-void ensure_instances(Language language, size_t instances);
+void ensure_tesseract_instances(Language language, size_t instances);
 
 //  Clear all TesseractAPI instances for all languages. Used for cleanup or
 //  forcing re-initialization.
 //  This is not safe to call while in any OCR is still running!
-void clear_cache();
+void clear_tesseract_cache();
 
 
 

SerialPrograms/Source/CommonTools/OCR/OCR_Routines.cpp

@@ -7,21 +7,68 @@
 #include "CommonFramework/ImageTypes/ImageRGB32.h"
 #include "CommonFramework/Tools/GlobalThreadPools.h"
 #include "CommonFramework/GlobalSettingsPanel.h"
+#include "CommonFramework/Exceptions/OperationFailedException.h"
 #include "CommonTools/Images/ImageFilter.h"
 #include "OCR_RawPaddleOCR.h"
-#include "OCR_RawOCR.h"
+#include "OCR_RawTesseractOCR.h"
 #include "OCR_DictionaryMatcher.h"
 #include "OCR_Routines.h"
 
-// #include <iostream>
-// using std::cout;
-// using std::endl;
+#include <iostream>
+using std::cout;
+using std::endl;
 
 
 
 namespace PokemonAutomation{
 namespace OCR{
 
+// return true if the user enabled PaddleOCR and its resource has been downloaded
+// NOTE: we are making the assumption that if English is downloaded, 
+// that all the other languages are downloaded too, since they're all downloaded
+// as a bundle.
+bool use_paddle_ocr(){
+    static bool use_paddle_ocr = GlobalSettings::instance().USE_PADDLE_OCR && paddle_ocr_language_available(Language::English);
+
+    return use_paddle_ocr;
+}
+
+bool ocr_language_available(Language language){
+    if (use_paddle_ocr()){
+        return  OCR::paddle_ocr_language_available(language);
+    }else{
+        return OCR::tesseract_language_available(language);
+    }
+}
+
+
+std::string ocr_read(Language language, const ImageViewRGB32& image, PageSegMode psm){
+    std::string ocr_text = "";
+    if (use_paddle_ocr()){
+        ocr_text = OCR::paddle_ocr_read(language, image);
+    }else{
+        ocr_text = OCR::tesseract_ocr_read(language, image, OCR::PageSegMode::SINGLE_LINE);
+    }
+    return ocr_text;
+}
+
+void ensure_ocr_instances(Language language, size_t instances){
+    if (use_paddle_ocr()){
+        OCR::ensure_paddle_ocr_instance(language);
+    }else{
+        OCR::ensure_tesseract_instances(language, instances);
+    }
+}
+
+void clear_ocr_cache(){
+    if (use_paddle_ocr()){
+        OCR::clear_paddle_ocr_cache();
+    }else{
+        OCR::clear_tesseract_cache();
+    }
+}
+
+
 
 StringMatchResult multifiltered_OCR(
     Language language, const DictionaryMatcher& dictionary, const ImageViewRGB32& image,
@@ -44,21 +91,14 @@ StringMatchResult multifiltered_OCR(
 
     double pixels_inv = 1. / (image.width() * image.height());
 
-    bool use_paddle_ocr = GlobalSettings::instance().USE_PADDLE_OCR;
-
     //  Run all the filters.
     SpinLock lock;
     StringMatchResult ret;
     GlobalThreadPools::computation_normal().run_in_parallel(
         [&](size_t index){
             const std::pair<ImageRGB32, size_t>& filtered = filtered_images[index];
 
-            std::string text;
-            if (use_paddle_ocr){
-                text = paddle_ocr_read(language, filtered.first);
-            }else{
-                text = ocr_read(language, filtered.first, psm);
-            }
+            std::string text = OCR::ocr_read(language, filtered.first, psm);
 
             // cout << "multifiltered_OCR: " << index << " -> " << text << endl;
             // filtered.first.save("test_" + std::to_string(index) + ".png");
@@ -110,12 +150,7 @@ StringMatchResult dictionary_OCR(
     }
 
     //  Run all the filters.
-    std::string text;
-    if (GlobalSettings::instance().USE_PADDLE_OCR){
-        text = paddle_ocr_read(language, image);
-    }else{
-        text = ocr_read(language, image, psm);
-    }
+    std::string text = OCR::ocr_read(language, image, psm);
 
     // cout << "dictionary_OCR: " << text << endl;
     // image.save("test_dictionary_OCR.png");

SerialPrograms/Source/CommonTools/OCR/OCR_Routines.h

@@ -10,7 +10,7 @@
 #include <cstdint>
 #include <vector>
 #include "CommonFramework/Language.h"
-#include "OCR_RawOCR.h"
+#include "OCR_RawTesseractOCR.h"
 
 namespace PokemonAutomation{
     class ImageViewRGB32;
@@ -30,9 +30,16 @@ struct TextColorRange{
     {}
 };
 
+bool ocr_language_available(Language language);
+
+std::string ocr_read(Language language, const ImageViewRGB32& image, PageSegMode psm = PageSegMode::SINGLE_LINE);
+
+void ensure_ocr_instances(Language language, size_t instances = 1);
+
+void clear_ocr_cache();
 
 // psm: Tesseract Page Segmentation mode. See
-//   SerialPrograms/Source/CommonTools/OCR/OCR_RawOCR.h:PageSegMode
+//   SerialPrograms/Source/CommonTools/OCR/OCR_RawTesseractOCR.h:PageSegMode
 StringMatchResult multifiltered_OCR(
     Language language, const DictionaryMatcher& dictionary, const ImageViewRGB32& image,
     const std::vector<TextColorRange>& text_color_ranges,

SerialPrograms/Source/CommonTools/OCR/OCR_TextMatcher.h

@@ -11,7 +11,6 @@
 #include <set>
 #include <map>
 #include <QString>
-#include "OCR_RawOCR.h"
 #include "OCR_StringMatchResult.h"
 
 namespace PokemonAutomation{