Re-attempt 10 times, some tweaks to OCR, pre-filter card pool HEAD main

5 files changed, 202 insertions, 151 deletions

@@ -17,6 +17,7 @@ include_directories(
   vendor/asio/asio/include
   vendor/base64/include
   ${tesseract_INCLUDE_DIRS}
+  #/Users/hatkirby/Repos/tesseract/include
   ${lept_INCLUDE_DIRS}
   ${GraphicsMagick_INCLUDE_DIRS}
   ${curlcpp_INCLUDE_DIRS}
@@ -26,6 +27,7 @@ include_directories(
 
 link_directories(
   ${tesseract_LIBRARY_DIRS}
+  #/Users/hatkirby/Repos/tesseract/build/here/usr/local/lib
   ${lept_LIBRARY_DIRS}
   ${GraphicsMagick_LIBRARY_DIRS}
   ${curlcpp_LIBRARY_DIRS}
@@ -0,0 +1,41 @@
+require 'json'
+
+input = File.open(ARGV[0]) do |file|
+  JSON.load(file)
+end
+
+filtered = input.select do |cardJson|
+  cardJson["object"] == "card" &&
+  # It needs to have a downloadable image
+  cardJson.count("image_uris") &&
+  # Make sure we can support the card layout
+  ["normal", "leveler", "saga"].include?(cardJson["layout"]) &&
+  # We only support modern and m15 frames
+  ["2015", "2003"].include?(cardJson["frame"]) &&
+  # Digital cards look slightly different so ignore them
+  !cardJson["digital"] &&
+  # Only use english printings
+  cardJson["lang"] == "en" &&
+  # Currently not supporting silver bordered cards
+  cardJson["border_color"] != "silver" &&
+  # It is hard to read the name of a planeswalker
+  !cardJson["type_line"].include?("Planeswalker") &&
+  # This cuts out checklists and special tokens
+  cardJson["type_line"] != "Card" &&
+  # Amonkhet invocations are impossible
+  cardJson["set"] != "mp2" &&
+  # Unknown Event is not a real thing huh
+  cardJson["set"] != "da1"
+end.map do |cardJson|
+  {
+    "id" => cardJson["id"],
+    "name" => cardJson["name"],
+    "imageUri" => cardJson["image_uris"]["png"],
+    "frame" => cardJson["frame"],
+    "artist" => cardJson["artist"]
+  }
+end
+
+File.open(ARGV[1], 'w') do |file|
+  JSON.dump(filtered, file)
+end
@@ -14,53 +14,28 @@ cardset::cardset(std::string filename) {
   nlohmann::json cardsJson = nlohmann::json::parse(contents.str());
 
   for (const auto& cardJson : cardsJson) {
-    if (
-        // The object needs to be a card
-        cardJson["object"] == "card" &&
-        // It needs to have a downloadable image
-        cardJson.count("image_uris") &&
-        // Make sure we can support the card layout
-        (cardJson["layout"] == "normal" || cardJson["layout"] == "leveler" ||
-         cardJson["layout"] == "saga") &&
-        // Digital cards look slightly different so ignore them
-        !cardJson["digital"] &&
-        // Only use english printings
-        cardJson["lang"] == "en" &&
-        // Currently not supporting silver bordered cards
-        cardJson["border_color"] != "silver" &&
-        // It is hard to read the name of a planeswalker
-        cardJson["type_line"].get<std::string>().find("Planeswalker") ==
-            std::string::npos &&
-        // This cuts out checklists and special tokens
-        cardJson["type_line"] != "Card" &&
-        // Amonkhet invocations are impossible
-        cardJson["set"] != "mp2" &&
-        // Unknown Event is not a real thing huh
-        cardJson["set"] != "da1" &&
-        // Ignore cards with the special legendary flare
-        (!cardJson.count("frame_effects") ||
-         !cardJson["frame_effects"].count("legendary"))) {
-      card_frame frame;
-
-      if (cardJson["frame"] == "2015") {
-        frame = card_frame::m2015;
-      } else if (cardJson["frame"] == "2003") {
-        frame = card_frame::modern;
-      } else {
-        continue;
-      }
+    // We're pre-processing the cardset now to reduce the work done here (but
+    // mostly the space taken by the data).
+    card_frame frame;
+
+    if (cardJson["frame"] == "2015") {
+      frame = card_frame::m2015;
+    } else if (cardJson["frame"] == "2003") {
+      frame = card_frame::modern;
+    } else {
+      continue;
+    }
 
-      size_t cardId = cards_.size();
-      cards_.emplace_back(cardId, cardJson["name"],
-                          cardJson["image_uris"]["png"], frame, cardJson["id"]);
+    size_t cardId = cards_.size();
+    cards_.emplace_back(cardId, cardJson["name"], cardJson["imageUri"], frame,
+                        cardJson["id"], cardJson["artist"]);
 
-      std::string canon = hatkirby::lowercase(cardJson["name"]);
+    std::string canon = hatkirby::lowercase(cardJson["name"]);
 
-      for (int i = 0; i < canon.length(); i++) {
-        titles_.add(canon, {cardId, i}, i);
+    for (int i = 0; i < canon.length(); i++) {
+      titles_.add(canon, {cardId, i}, i);
 
-        chars_.insert(canon.at(i));
-      }
+      chars_.insert(canon.at(i));
     }
   }
 }
@@ -17,14 +17,16 @@ struct card {
   std::string imageUri;
   card_frame frame;
   std::string uuid;
+  std::string artist;
 
   card(size_t id, std::string name, std::string imageUri, card_frame frame,
-       std::string uuid)
+       std::string uuid, std::string artist)
       : id(id),
         name(std::move(name)),
         imageUri(std::move(imageUri)),
         frame(frame),
-        uuid(std::move(uuid)) {}
+        uuid(std::move(uuid)),
+        artist(std::move(artist)) {}
 };
 
 class cardset {
@@ -55,6 +55,24 @@ void wizard::set_status_callback(status_callback_type callback) {
 }
 
 Magick::Image wizard::run() {
+  tesseract_ptr api{new tesseract::TessBaseAPI()};
+
+  if (api->Init(nullptr, "eng", tesseract::OEM_TESSERACT_LSTM_COMBINED)) {
+    throw std::runtime_error("Could not initialize tesseract");
+  }
+
+  std::string allChars;
+  for (char ch : cards_.getCharacters()) {
+    allChars.push_back(ch);
+    if (std::isalpha(ch)) {
+      allChars.push_back(std::toupper(ch));
+    }
+  }
+
+  if (!api->SetVariable("tessedit_char_whitelist", allChars.c_str())) {
+    throw std::runtime_error("Could not set tesseract allowlist");
+  }
+
   std::string text = text_;  //"what the heck, it's just some gay guy";
   // getline(std::cin, text);
   if (text.back() == '\r') {
@@ -68,104 +86,109 @@ Magick::Image wizard::run() {
 
   std::string canonText = hatkirby::lowercase(text);
   designer des(canonText, cards_.getTitles());
-  std::list<usage> res = des.generate(rng_);
 
   Magick::Image endImage;
-  bool firstSlice = false;
-  int ui = 0;
-
-  for (const usage& u : res) {
-    const card& theCard = cards_.getCard(u.cardId);
-    const std::string& cardName = theCard.name;
-
-    std::cout << cardName.substr(0, u.strIndex) << "["
-              << cardName.substr(u.strIndex, u.strLen) << "]"
-              << cardName.substr(u.strIndex + u.strLen) << std::endl;
-
-    std::cout << "Downloading image..." << std::endl;
-    if (status_callback_) {
-      status_callback_("Downloading image...");
-    }
-
-    Magick::Image cardImg = images_.get(theCard.uuid, theCard.imageUri);
-
-    std::cout << "Reading text..." << std::endl;
-    if (status_callback_) {
-      status_callback_("Reading text...");
-    }
-
-    Magick::Image titleImg = cardImg;
-    titleImg.magick("TIFF");
-
-    // titleImg.threshold(MaxRGB / 2);
-    titleImg.write("pre.tif");
-
-    Magick::Geometry margin;
-
-    if (theCard.frame == card_frame::m2015) {
-      margin = Magick::Geometry{595, 46, 57, 54};
-    } else if (theCard.frame == card_frame::modern) {
-      margin = Magick::Geometry{581, 50, 63, 57};
-    }
-
-    titleImg.crop(margin);
-    titleImg.zoom({margin.width() * 5, margin.height() * 5});
-
-    // titleImg.quantizeColorSpace(Magick::GRAYColorspace);
-    // titleImg.quantizeColors(2);
-    // titleImg.quantize();
-    titleImg.backgroundColor("white");
-    titleImg.matte(false);
-    titleImg.resolutionUnits(Magick::PixelsPerInchResolution);
-    titleImg.density({300, 300});
-    titleImg.type(Magick::GrayscaleType);
+  bool success = false;
+  for (int attempt = 0; attempt < 10; attempt++) {
+    endImage = {};
+
+    std::list<usage> res = des.generate(rng_);
+
+    bool firstSlice = false;
+    int ui = 0;
+
+    bool failure = false;
+    for (const usage& u : res) {
+      const card& theCard = cards_.getCard(u.cardId);
+      const std::string& cardName = theCard.name;
+
+      std::cout << cardName.substr(0, u.strIndex) << "["
+                << cardName.substr(u.strIndex, u.strLen) << "]"
+                << cardName.substr(u.strIndex + u.strLen) << std::endl;
+
+      std::cout << "Downloading image..." << std::endl;
+      if (status_callback_) {
+        std::ostringstream status_stream;
+        status_stream << "Processing card " << (ui + 1) << "/" << res.size()
+                      << " (attempt " << (attempt + 1) << ")...";
+        status_callback_(status_stream.str());
+      }
 
-    titleImg.write("title.tif");
+      Magick::Image cardImg = images_.get(theCard.uuid, theCard.imageUri);
 
-    Magick::Blob titleBlob;
-    titleImg.write(&titleBlob);
+      std::cout << "Reading text..." << std::endl;
 
-    pix_ptr titlePix{
-        pixReadMemTiff(reinterpret_cast<const unsigned char*>(titleBlob.data()),
-                       titleBlob.length(), 0)};
+      Magick::Image titleImg = cardImg;
+      titleImg.magick("TIFF");
 
-    tesseract_ptr api{new tesseract::TessBaseAPI()};
+      // titleImg.threshold(MaxRGB / 2);
+      titleImg.write("pre.tif");
 
-    if (api->Init(nullptr, "eng")) {
-      throw std::runtime_error("Could not initialize tesseract");
-    }
+      Magick::Geometry margin;
 
-    api->SetImage(titlePix.get());
-    api->Recognize(nullptr);
+      if (theCard.frame == card_frame::m2015) {
+        margin = Magick::Geometry{595, 50, 57, 54};
+      } else if (theCard.frame == card_frame::modern) {
+        margin = Magick::Geometry{581, 50, 63, 57};
+      }
 
-    tesseract::ResultIterator* ri = api->GetIterator();
-    tesseract::PageIteratorLevel level = tesseract::RIL_TEXTLINE;
-    bool foundName = false;
-    size_t extraChars = 0;
+      titleImg.crop(margin);
+      titleImg.zoom({margin.width() * 5, margin.height() * 5});
+
+      // titleImg.quantizeColorSpace(Magick::GRAYColorspace);
+      // titleImg.quantizeColors(2);
+      // titleImg.quantize();
+      titleImg.backgroundColor("white");
+      titleImg.matte(false);
+      titleImg.resolutionUnits(Magick::PixelsPerInchResolution);
+      titleImg.density({300, 300});
+      titleImg.type(Magick::GrayscaleType);
+      titleImg.gamma(3);
+      // titleImg.level(0.2, 0.8);
+
+      titleImg.write("title.tif");
+
+      Magick::Blob titleBlob;
+      titleImg.write(&titleBlob);
+
+      pix_ptr titlePix{pixReadMemTiff(
+          reinterpret_cast<const unsigned char*>(titleBlob.data()),
+          titleBlob.length(), 0)};
+
+      api->SetImage(titlePix.get());
+      api->Recognize(nullptr);
+
+      tesseract::ResultIterator* ri = api->GetIterator();
+      tesseract::PageIteratorLevel level = tesseract::RIL_TEXTLINE;
+      bool foundName = false;
+      size_t extraChars = 0;
+
+      if (ri) {
+        do {
+          const char* line = ri->GetUTF8Text(level);
+
+          if (line) {
+            std::string lineStr(line);
+
+            size_t leadin = hatkirby::lowercase(lineStr).find(
+                hatkirby::lowercase((cardName)));
+            if (leadin != std::string::npos) {
+              foundName = true;
+              extraChars = leadin;
+
+              break;
+            } /* else {
+                std::cout << "WRONG: " << lineStr << std::endl;
+              }*/
+          }
+        } while (ri->Next(level));
+      }
 
-    if (ri) {
-      do {
-        const char* line = ri->GetUTF8Text(level);
-
-        if (line) {
-          std::string lineStr(line);
-
-          size_t leadin = hatkirby::lowercase(lineStr).find(
-              hatkirby::lowercase((cardName)));
-          // if (leadin != std::string::npos)
-          {
-            foundName = true;
-            // extraChars = leadin;
-
-            break;
-          } /* else {
-             std::cout << "WRONG: " << lineStr << std::endl;
-           }*/
-        }
-      } while (ri->Next(level));
-    }
+      if (!foundName) {
+        failure = true;
+        break;
+      }
 
-    if (foundName) {
       level = tesseract::RIL_SYMBOL;
 
       for (int i = 0; i < extraChars; i++) {
@@ -233,32 +256,40 @@ Magick::Image wizard::run() {
 
       cardImg.magick("PNG");
       cardImg.write("slice.png");
-    } else {
-      std::cout << "Didn't find name" << std::endl;
-    }
 
-    if (!firstSlice) {
-      firstSlice = true;
-      endImage = cardImg;
+      if (!firstSlice) {
+        firstSlice = true;
+        endImage = cardImg;
 
-      if (theCard.frame == card_frame::m2015) {
-        // endImage.extent({endImage.columns(), endImage.rows() + 6},
-        // Magick::SouthGravity);
-      }
-    } else {
-      int xoff = endImage.columns();
+        if (theCard.frame == card_frame::m2015) {
+          // endImage.extent({endImage.columns(), endImage.rows() + 6},
+          // Magick::SouthGravity);
+        }
+      } else {
+        int xoff = endImage.columns();
 
-      endImage.backgroundColor("black");
+        endImage.backgroundColor("black");
 
-      endImage.extent({endImage.columns() + cardImg.columns(), cardImg.rows()},
-                      Magick::WestGravity);
+        endImage.extent(
+            {endImage.columns() + cardImg.columns(), cardImg.rows()},
+            Magick::WestGravity);
 
-      endImage.composite(cardImg, xoff,
-                         (theCard.frame == card_frame::m2015) ? 6 : 0);
+        endImage.composite(cardImg, xoff,
+                           (theCard.frame == card_frame::m2015) ? 6 : 0);
+      }
+
+      // break;
+      ui++;
     }
 
-    // break;
-    ui++;
+    if (!failure) {
+      success = true;
+      break;
+    }
+  }
+
+  if (!success) {
+    throw std::domain_error("Could not generate card");
   }
 
   endImage.magick("PNG");