Added Doxygen documentation, reverted detect change, improved detection pipeline

Changes:
- Added Doxygen documentation for marco_annunziata.cpp and marco_annunziata.hpp
- Updated README with doxygen documentation and instructions
- Reverted detect change made in commit 4ad8bae, solving related bug on small images
- Improved detection pipeline for tiled images by doing a final detection on the whole image (solved bug on 5C0.jpg)
- Added clearDetections()
- Changed DrawBoundingBoxes to have less parameters (more difficult to pass wrong arguments)

Author: Remdox

.gitignore

@@ -6,3 +6,4 @@ output/*
 data/test/*
 build/*
 external/onnxruntime-linux-x64-gpu-1.21.0.tgz
+doxygen_docs/

README.md

@@ -126,6 +126,11 @@ In order to do this, some tools like [CVAT](https://www.cvat.ai/), [Label studio
 
 # Code
 ## Doxygen Documentation
+The [pdf file](./doxygen_docs/latex/refman.pdf) of the Doxygen documentation is already bundled inside the repo.
+To get the html version (or to build the pdf from scratch), the Doxygen documentation can be generated by running:
+```
+doxygen doxygen.txt
+```
 
 ## Object Detection and Initial Classification
 ### Training of the model and exporting to ONNX format
@@ -167,3 +172,4 @@ For custom files, ground truths can be manually added inside the `data/test/grou
 If the ground truths for a custom file are not provided, the program will still run the detection pipeline without computing metrics.
 
 # Output
+The output of the detection is saved in a output/ folder.