fixing ocr number multifilter

Author: Gin

SerialPrograms/Source/CommonTools/OCR/OCR_NumberReader.cpp

@@ -101,43 +101,21 @@ int read_number(Logger& logger, const ImageViewRGB32& image, Language language){
     return number;
 }
 
-int read_number_waterfill(
-    Logger& logger, const ImageViewRGB32& image,
-    uint32_t rgb32_min, uint32_t rgb32_max,    
-    bool text_inside_range,
-    int8_t line_index
-){
-    std::string ocr_text = read_number_waterfill_no_normalization(
-        logger,
-        image,
-        rgb32_min, rgb32_max,
-        text_inside_range
-    );
-
-    std::string normalized = run_number_normalization(ocr_text);
-
-    std::string line_index_str = "";
-    if (line_index != -1){
-        line_index_str = "Line " + std::to_string(line_index) + ": ";
-    }
-    if (normalized.empty()){
-        logger.log(line_index_str + "OCR Text: \"" + ocr_text + "\" -> \"" + normalized + "\" -> Unable to read.", COLOR_RED);
-        return -1;
-    }
-
-    int number = std::atoi(normalized.c_str());
-    logger.log(line_index_str + "OCR Text: \"" + ocr_text + "\" -> \"" + normalized + "\" -> " + std::to_string(number));
-
-    return number;
-}
-
 
+// Run OCR on each individual character in the string of numbers.
+// Return empty string if OCR fails.
+//
+// text_inside_range: binary filter is applied to the image so that any pixels within the color range will be turned black, and everything else will be white
+// width_max: return empty string if any character's width is greater than width_max (likely means that two characters are touching, and so are treated as one large character)
+// min_digit_area: if a character has area (aka pixel count) smaller than this value (likely noise or punctuations), skip this character
+// check_empty_string: if set to true, return empty string (and stop evaluation) if any character returns an empty string from OCR
 std::string read_number_waterfill_no_normalization(
     Logger& logger, const ImageViewRGB32& image,
     uint32_t rgb32_min, uint32_t rgb32_max,    
-    bool text_inside_range,
-    size_t width_max,
-    bool check_empty_string
+    bool text_inside_range = true,
+    size_t width_max = (size_t)-1,
+    size_t min_digit_area = 20,
+    bool check_empty_string = false
 ){
     using namespace Kernels::Waterfill;
 
@@ -159,7 +137,7 @@ std::string read_number_waterfill_no_normalization(
     std::map<size_t, WaterfillObject> map;
     {
         std::unique_ptr<WaterfillSession> session = make_WaterfillSession(matrix);
-        auto iter = session->make_iterator(20);
+        auto iter = session->make_iterator(min_digit_area);
         WaterfillObject object;
         while (map.size() < 16 && iter->find_next(object, true)){
             if (object.width() > width_max){
@@ -205,12 +183,45 @@ bool is_digits(const std::string &str)
     return std::all_of(str.begin(), str.end(), ::isdigit);
 }
 
+
+int read_number_waterfill(
+    Logger& logger, const ImageViewRGB32& image,
+    uint32_t rgb32_min, uint32_t rgb32_max,    
+    bool text_inside_range,
+    int8_t line_index
+){
+    std::string ocr_text = read_number_waterfill_no_normalization(
+        logger,
+        image,
+        rgb32_min, rgb32_max,
+        text_inside_range
+    );
+
+    std::string normalized = run_number_normalization(ocr_text);
+
+    std::string line_index_str = "";
+    if (line_index != -1){
+        line_index_str = "Line " + std::to_string(line_index) + ": ";
+    }
+    if (normalized.empty()){
+        logger.log(line_index_str + "OCR Text: \"" + ocr_text + "\" -> \"" + normalized + "\" -> Unable to read.", COLOR_RED);
+        return -1;
+    }
+
+    int number = std::atoi(normalized.c_str());
+    logger.log(line_index_str + "OCR Text: \"" + ocr_text + "\" -> \"" + normalized + "\" -> " + std::to_string(number));
+
+    return number;
+}
+
+
 int read_number_waterfill_multifilter(
     Logger& logger, const ImageViewRGB32& image,
     std::vector<std::pair<uint32_t, uint32_t>> filters,    
-    size_t width_max,
     bool text_inside_range,
     bool prioritize_numeric_only_results, 
+    size_t width_max,
+    size_t min_digit_area,
     int8_t line_index
 ){
     std::string line_index_str = "";
@@ -226,13 +237,14 @@ int read_number_waterfill_multifilter(
 
             uint32_t rgb32_min = filter.first;
             uint32_t rgb32_max = filter.second;
-            bool check_empty_string = false;
+            bool check_empty_string = true;
             std::string ocr_text = read_number_waterfill_no_normalization(
                 logger,
                 image,
                 rgb32_min, rgb32_max,
                 text_inside_range,
                 width_max,
+                min_digit_area,
                 check_empty_string
             );
 

SerialPrograms/Source/CommonTools/OCR/OCR_NumberReader.h

@@ -36,35 +36,25 @@ int read_number_waterfill(
     int8_t line_index = -1
  );
 
-// run OCR on each individual character in the string of numbers.
-// return empty string if OCR fails
-//
-// text_inside_range: binary filter is applied to the image so that any pixels within the color range will be turned black, and everything else will be white
-// width_max: return empty string if any character's width is greater than width_max (likely means that two characters are touching, and so are treated as one large character)
-// check_empty_string: if set to true, return empty string (and stop evaluation) if any character returns an empty string from OCR
- std::string read_number_waterfill_no_normalization(
-    Logger& logger, const ImageViewRGB32& image,
-    uint32_t rgb32_min, uint32_t rgb32_max,
-    bool text_inside_range = true,
-    size_t width_max = (size_t)-1,
-    bool check_empty_string = false
- );
 
 // Try OCR with all the given color filters. still running OCR on each individual character
 // Return the best majority candidate. return -1 if failed to read.
+// width_max: if a character width is greater than this value (likely means two characters are touching, will cause bad read), skip this filter.
+// min_digit_area: if a character has area (aka pixel count) smaller than this value (likely noise or punctuations), skip this character
 // 
 // prioritize_numeric_only_results: 
 //  - if true: if OCR reads only numeric characters, the candidate gets 2 votes. If OCR reads non-numeric characters, the candidate gets only 1 vote.
 //  - if false: all reads only get 1 vote
 //
-// line_index: specifies the current number's row. for logging purposes, when multithreaded.
+// log_line_index: adds an index prefix to the logging lines for logging purposes when calling this function in parallel.
 int read_number_waterfill_multifilter(
     Logger& logger, const ImageViewRGB32& image,
     std::vector<std::pair<uint32_t, uint32_t>> filters,
-    size_t width_max = (size_t)-1,
     bool text_inside_range = true,
     bool prioritize_numeric_only_results = true,
-    int8_t line_index = -1
+    size_t width_max = (size_t)-1,
+    size_t min_digit_area = 20,
+    int8_t log_line_index = -1
  );
 
 

SerialPrograms/Source/PokemonLZA/Inference/Boxes/PokemonLZA_BoxInfoDetector.cpp

@@ -205,19 +205,32 @@ void BoxDexNumberDetector::make_overlays(VideoOverlaySet& items) const{
 bool BoxDexNumberDetector::detect(const ImageViewRGB32& screen){
     const size_t max_dex_number = std::max(LUMIOSE_DEX_SLUGS().size(), HYPERSPACE_DEX_SLUGS().size());
 
-    const ImageViewRGB32 dex_image_crop = extract_box_reference(screen, m_dex_number_box);
     // const bool in_range_black = false;
     // const ImageRGB32 black_white_dex_image_crop = to_blackwhite_rgb32_range(dex_image_crop, in_range_black, 0xff808080, 0xffffffff);
     // black_white_dex_image_crop.save("blackwhite_number.png");
     // const int dex_number = OCR::read_number(m_logger, black_white_dex_image_crop);
     // const int dex_number = OCR::read_number_waterfill(m_logger, dex_image_crop, 0xff808080, 0xffffffff, false);
-    const int dex_number = OCR::read_number_waterfill_multifilter(m_logger, dex_image_crop, {
-        {0x0, 0xff202020},
-        {0x0, 0xff404040},
-        {0x0, 0xff606060},
-        {0x0, 0xff808080},
-        {0x0, 0xffA0A0A0},
-    });
+    
+    const int dex_number = [&](){
+        const ImageViewRGB32 dex_image_crop = extract_box_reference(screen, m_dex_number_box);
+        const bool text_inside_range = true;
+        const bool prioritize_numeric_only_results = true;
+        const size_t width_max = SIZE_MAX;
+        // To accomodate the dex number "No. xxx" for all language, we have to make the dex number crop to cover the "dot" character
+        // for some languages. We have to use `min_digit_area` to filter out the dot when doing OCR.
+        // The min digit area computation is that any dot with size smaller than image_crop.height()/5 is filtered out when OCR.
+        const size_t min_digit_area = dex_image_crop.height()*dex_image_crop.height() / 25;
+        return OCR::read_number_waterfill_multifilter(m_logger, dex_image_crop,
+            {
+                {0x0, 0xff202020},
+                {0x0, 0xff404040},
+                {0x0, 0xff606060},
+                {0x0, 0xff808080},
+                {0x0, 0xffA0A0A0},
+            },
+            text_inside_range, prioritize_numeric_only_results, width_max, min_digit_area
+        );
+    }();
     if (dex_number <= 0 || dex_number > static_cast<int>(max_dex_number)) {
         m_dex_number = 0;
         m_dex_number_when_error = dex_number;

SerialPrograms/Source/PokemonSV/Inference/ItemPrinter/PokemonSV_ItemPrinterMaterialDetector.cpp

@@ -136,13 +136,16 @@ int16_t ItemPrinterMaterialDetector::read_number(
             }
         }();
 
-    size_t max_width = (size_t)((double)24 * screen.width() / 1080);
-
+    const bool text_inside_range = true;
+    const bool prioritize_numeric_only_results = true;
+    const size_t max_width = (size_t)((double)24 * screen.width() / 1080);
+    const size_t min_digit_area = 20;
     int16_t number = (int16_t)OCR::read_number_waterfill_multifilter(
         logger,
         cropped, filters,
+        text_inside_range, prioritize_numeric_only_results,
         max_width,
-        true, true,
+        min_digit_area,
         row_index
     );