Parallel ocr evaluation for sections of screen + many refactorings

3 files changed, 83 insertions, 38 deletions

diff --git a/include/image.h b/include/image.h
new file mode 100644
index 0000000..ecbc255
--- /dev/null
+++ b/include/image.h
@@ -0,0 +1,19 @@
+#pragma once
+#include <leptonica/allheaders.h>
+
+namespace image {
+// Preprocessing configuration
+static const float contrast = 0.3;
+static const float sharpness = 0.7;
+static const float scaleFactor = 1;
+static const float grayscaleWeightRed = 0.114;
+static const float grayscaleWeightGreen = 0.587;
+static const float grayscaleWeightBlue = 0.299;
+
+Pix *loadImage(const char *imagePath);
+} // namespace image
+
+#define INLINE_IMAGE_PROC(process)                                             \
+  temp = process;                                                              \
+  pixDestroy(image);                                                           \
+  *image = temp;
diff --git a/include/libchelleport.h b/include/libchelleport.h
index e6a074d..b69466e 100644
--- a/include/libchelleport.h
+++ b/include/libchelleport.h
@@ -1,46 +1,12 @@
+#pragma once
 #include <chrono>
+#include <iostream>
 #include <leptonica/allheaders.h>
+#include <memory>
 #include <tesseract/baseapi.h>
 #include <vector>
 
-// NOTE: Remember to update size and alignment in ocr hs module on change
-struct OCRMatch {
-  int startX, startY;
-  int endX, endY;
-  const char *text;
-};
-
-// OCR configuration
-#define CONFIDENCE_THRESHOLD 25.
-#define MIN_CHARACTER_COUNT 3
-const tesseract::PageIteratorLevel RESULT_ITER_MODE = tesseract::RIL_WORD;
-
-// Preprocessing configuration
-const float contrast = 0.3;
-const float sharpness = 0.7;
-const float scaleFactor = 1;
-const float grayscaleWeightRed = 0.114;
-const float grayscaleWeightGreen = 0.587;
-const float grayscaleWeightBlue = 0.299;
-
-extern "C" {
-OCRMatch *findWordCoordinates(const char *image_path, /* returns */ int *size);
-}
-
-tesseract::TessBaseAPI *initializeTesseract();
-
-Pix *loadImage(const char *imagePath);
-
-std::vector<OCRMatch> extractTextCoordinates(const char *imagePath);
-
-void printMatch(const OCRMatch &match);
-
-void preprocessImage(Pix **image);
-
-#define INLINE_IMAGE_PROC(process)                                             \
-  temp = process;                                                              \
-  pixDestroy(image);                                                           \
-  *image = temp;
+#include "./recognizer.h"
 
 #define MEASURE(label, stmts)                                                  \
   auto start = std::chrono::high_resolution_clock::now();                      \
@@ -49,3 +15,13 @@ void preprocessImage(Pix **image);
   auto duration =                                                              \
       std::chrono::duration_cast<std::chrono::microseconds>(end - start);      \
   std::cout << label << ": " << duration.count() / 1000.0 << " ms" << std::endl;
+
+extern "C" {
+OCRMatch *findWordCoordinates(const char *image_path, /* returns */ int *size);
+}
+
+std::vector<OCRMatch> extractTextMatches(const char *imagePath);
+
+std::vector<OCRMatch>
+runRecognizers(std::vector<std::unique_ptr<Recognizer>> &recognizers,
+               Pix *image);
diff --git a/include/recognizer.h b/include/recognizer.h
new file mode 100644
index 0000000..57747bb
--- /dev/null
+++ b/include/recognizer.h
@@ -0,0 +1,50 @@
+#pragma once
+#include <iostream>
+#include <leptonica/allheaders.h>
+#include <tesseract/baseapi.h>
+
+#include "./image.h"
+
+// OCR configuration
+#define CONFIDENCE_THRESHOLD 25.
+#define MIN_CHARACTER_COUNT 3
+const tesseract::PageIteratorLevel ITER_LEVEL = tesseract::RIL_WORD;
+
+// NOTE: Remember to update size and alignment in ocr hs module on change
+struct OCRMatch {
+  int startX, startY;
+  int endX, endY;
+  const char *text;
+};
+
+class Recognizer {
+  tesseract::TessBaseAPI *tesseract;
+  int x, y, width, height;
+  bool failed = false;
+
+public:
+  Recognizer(int x, int y, int width, int height)
+      : x(x), y(y), width(width), height(height) {
+    initializeTesseract();
+  }
+
+  ~Recognizer() { tesseract->End(); }
+
+  void fail(const char *msg);
+
+  void recognize(Pix *image);
+
+  OCRMatch *fetchMatch(tesseract::ResultIterator *iterator);
+
+  std::vector<OCRMatch> getResults();
+
+private:
+  void initializeTesseract();
+};
+
+inline void printMatch(const OCRMatch &match) {
+  std::cout << "Text: " << match.text << "; Position: (" << match.startX << ","
+            << match.startY << ") -> (" << match.endX << "," << match.endY
+            << ")" << std::endl
+            << std::endl;
+}