Merge pull request #7 from electronicarts/dev-Ricardo/FixLineAproximationError

Fix line approximation runtime error

Author: alexlentini

src/Log.cpp

@@ -14,7 +14,7 @@ namespace tik
 		m_CoreLogger = std::make_shared<spdlog::logger>("CoreLogger");
 
 		// Then the logging level can be set with the following function
-		cv::utils::logging::setLogLevel(cv::utils::logging::LogLevel::LOG_LEVEL_SILENT);
+		
 
 		//add sinks to logger
 		if (console)

src/Textbox.cpp

@@ -20,11 +20,13 @@ namespace tik
 		int boxWidth = bottomRight.x - topLeft.x;
 
 		textBoxRect = cv::Rect(topLeft.x, topLeft.y, boxWidth, boxHeight);
+		textRect = textBoxRect;
 		textSubMat = frameImg(textBoxRect);
 	}
 
 	TextBox::TextBox(cv::Rect rect, cv::Mat frameImg) : textBoxRect(rect)
 	{
+		textRect = textBoxRect;
 		textSubMat = frameImg(textBoxRect);
 	}
 

src/TextboxDetectionDB.cpp

@@ -99,24 +99,38 @@ namespace tik {
 		if (boxes.empty()) return { mergedLines, boxes };
 
 		// Sort boxes by the top y-coordinate
-		auto sortedBoxes = boxes;
+		auto& sortedBoxes = boxes;
 		for (auto& textBox : sortedBoxes)
 		{
 			textBox.calculateTextBoxLuminance(sRGB_LUT);
 		}
+
 		for (auto& textBox : sortedBoxes)
 		{
 			textBox.calculateTextMask();
 			auto boxRect = textBox.getTextBoxRect();
 			auto textRect = textBox.getTextRect();
 
-			textBox = { cv::Rect{ boxRect.x + textRect.x - 1, boxRect.y + textRect.y - 1, textRect.width + 2, textRect.height + 2 },img };
+			int x = std::max(boxRect.x + textRect.x, 0);
+			int y = std::max(boxRect.y + textRect.y, 0);
+			int w = std::min(boxRect.width, (img.cols - x));
+			int h = std::min(boxRect.height, (img.rows - y));
+
+			auto tb = cv::Rect{ x , y, w, h };
+			textBox = { tb,img };
 		}
 
-		// boxRect.x + textRect.x - 1, boxRect.y + textRect.y - 1, textRect.width + 2, textRect.height + 2
+		// Sort vertically
 		std::sort(sortedBoxes.begin(), sortedBoxes.end(), [](const TextBox& a, const TextBox& b) {
 			return a.getTextBoxRect().y < b.getTextBoxRect().y;
-		});
+			});
+		// Sort horizontally after vertically
+		std::sort(sortedBoxes.begin(), sortedBoxes.end(), [&](const TextBox& a, const TextBox& b) {
+			if (std::abs(a.getTextBoxRect().y - b.getTextBoxRect().y) < MAX_Y_DIFF) {
+				return a.getTextBoxRect().x < b.getTextBoxRect().x;
+			}
+			return a.getTextBoxRect().y < b.getTextBoxRect().y;
+			});
 
 		TextBox currentLine = sortedBoxes[0];
 

src/TextboxDetectionEAST.cpp

@@ -125,7 +125,7 @@ namespace tik {
 		if (boxes.empty()) return { mergedLines, boxes };
 
 		// Sort boxes by the top y-coordinate
-		auto sortedBoxes = boxes;
+		auto& sortedBoxes = boxes;
 		for (auto& textBox : sortedBoxes)
 		{
 			textBox.calculateTextBoxLuminance(sRGB_LUT);
@@ -138,8 +138,8 @@ namespace tik {
 
 			int x = std::max(boxRect.x + textRect.x , 0);
 			int y = std::max(boxRect.y + textRect.y , 0);
-			int w = std::min(textRect.width  , (img.cols - x));
-			int h = std::min(textRect.height , (img.rows - y));
+			int w = std::min(boxRect.width  , (img.cols - x));
+			int h = std::min(boxRect.height , (img.rows - y));
 
 			auto tb = cv::Rect{x , y, w, h };
 			textBox = { tb,img };

vcpkg.json

@@ -8,7 +8,12 @@
     "spscqueue",
     "gtest",
     "spdlog",
-    "benchmark"
+    "benchmark",
+    {
+      "name": "ffmpeg",
+      "features": [ "openh264", "openjpeg", "zlib" ]
+    }
+
   ],
   "features": {
     "cuda-opencv": {