Added experimental serialisation to JSON for cv::Mat's

This is needed by the json-to-cereal-binary app to read the JSON generated from the Matlab BFM converter script.

include/eos/morphablemodel/io/mat_cerealisation.hpp

 /*
- * superviseddescent: A C++11 implementation of the supervised descent
- *                    optimisation method
- * File: superviseddescent/matcerealisation.hpp
+* Eos - A 3D Morphable Model fitting library written in modern C++11/14.
+*
+* File: include/eos/morphablemodel/io/mat_cerealisation.hpp
  *
- * Copyright 2015 Patrik Huber
+ * Copyright 2015, 2016 Patrik Huber
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -29,6 +29,11 @@
 /**
  * @brief Serialisation of OpenCV \c cv::Mat matrices for the serialisation
  * library cereal (http://uscilab.github.io/cereal/index.html).
+ *
+ * Contains serialisation for \c cv::Mat matrices to binary archives, and
+ * serialisation of cv::Vec2f.
+ *
+ * Contains also an experimental serialisation to save/load cv::Mat's from JSON.
  */
 namespace cv {
 
@@ -40,7 +45,8 @@ namespace cv {
  * @param[in] ar The archive to serialise to.
  * @param[in] mat The matrix to serialise.
  */
-template<class Archive>
+template<class Archive, cereal::traits::DisableIf<cereal::traits::is_text_archive<Archive>::value>
+	= cereal::traits::sfinae>
 void save(Archive& ar, const cv::Mat& mat)
 {
 	int rows, cols, type;
@@ -75,7 +81,8 @@ void save(Archive& ar, const cv::Mat& mat)
  * @param[in] ar The archive to deserialise from.
  * @param[in] mat The matrix to deserialise into.
  */
-template<class Archive>
+template<class Archive, cereal::traits::DisableIf<cereal::traits::is_text_archive<Archive>::value>
+	= cereal::traits::sfinae>
 void load(Archive& ar, cv::Mat& mat)
 {
 	int rows, cols, type;
@@ -99,6 +106,108 @@ void load(Archive& ar, cv::Mat& mat)
 	}
 };
 
+
+/**
+ * @brief Serialise a cv::Mat using cereal, for text archives (JSON specifically).
+ *
+ * Experimental: This specialisation is enabled for text archives (e.g. XML, JSON),
+ * and was created to convert the JSON output from the BFM Matlab converter script
+ * to a cereal binary model.
+ * 
+ * Notes:
+ * - Only for 2-dim matrices, and float values, i.e. CV_32FC1 cv::Mat's.
+ *   Actually, _only_ 32FC1, because of the load() method. In fact, maybe we should
+ *   store the type, since we're storing cv::Mat's.
+ * - Writes the data row-wise to a json array? or a json object?
+ *
+ * @param[in] ar The archive to serialise to.
+ * @param[in] mat The matrix to serialise.
+ */
+template <class Archive,
+	cereal::traits::EnableIf<cereal::traits::is_text_archive<Archive>::value>
+	= cereal::traits::sfinae>
+	void save(Archive& ar, const cv::Mat& mat)
+{
+	int rows, cols, type;
+	bool continuous;
+
+	rows = mat.rows;
+	cols = mat.cols;
+	type = mat.type();
+	continuous = mat.isContinuous();
+
+	//ar & rows & cols & type & continuous;
+	assert(mat.dims == 2); // correct?
+
+	if (continuous) { // We go row by row anyway so no need for this distinction?
+		//const int data_size = rows * cols * static_cast<int>(mat.elemSize());
+		//std::vector<float> test(mat.begin<float>(), mat.end<float>());
+		//ar & test;
+
+		std::vector<std::vector<float>> mat_data;
+		for (int i = 0; i < rows; i++) {
+			Mat this_row = mat.row(i); // need a temporary, otherwise goes up to 8GB RAM usage
+			mat_data.push_back(std::vector<float>(this_row.begin<float>(), this_row.end<float>()));
+		}
+		ar & cereal::make_nvp("data", mat_data); // Can we somehow not give this a name and make it like the "root" node, part of the parent object? Maybe look at the std::string serialisation?
+	}
+	else {
+		const int row_size = cols * static_cast<int>(mat.elemSize());
+		for (int i = 0; i < rows; i++) {
+			//auto row_data = cereal::binary_data(mat.ptr(i), row_size);
+			//ar & row_data;
+		}
+	}
+};
+
+
+/**
+ * @brief De-serialise a cv::Mat using cereal, for text archives (JSON specifically).
+ *
+ * Experimental: This specialisation is enabled for text archives (e.g. XML, JSON),
+ * and was created to convert the JSON output from the BFM Matlab converter script
+ * to a cereal binary model.
+ * See the notes of the save() method!
+ *
+ * @param[in] ar The archive to deserialise from.
+ * @param[in] mat The matrix to deserialise into.
+ */
+template <class Archive,
+	cereal::traits::EnableIf<cereal::traits::is_text_archive<Archive>::value>
+	= cereal::traits::sfinae>
+	void load(Archive& ar, cv::Mat& mat)
+{
+	//int rows, cols, type;
+	//bool continuous;
+
+	//ar & rows & cols & type & continuous;
+
+	std::vector<std::vector<float>> mat_data;
+	ar & mat_data;
+	assert(mat_data.size() > 0); // hmm can't store empty cv::Mat's... not so nice. Will create problems with SFM shape-only models?
+	int rows = static_cast<int>(mat_data.size());
+	int cols = static_cast<int>(mat_data[0].size());
+	mat.create(rows, cols, CV_32FC1);
+	for (int r = 0; r < rows; ++r) {
+		for (int c = 0; c < cols; ++c) {
+			mat.at<float>(r, c) = mat_data[r][c];
+		}
+	}
+
+/*	mat.create(rows, cols, type);
+	const int data_size = rows * cols * static_cast<int>(mat.elemSize());
+	auto mat_data = cereal::binary_data(mat.ptr(), data_size);
+	ar & mat_data;
+	
+	mat.create(rows, cols, type);
+	const int row_size = cols * static_cast<int>(mat.elemSize());
+	for (int i = 0; i < rows; i++) {
+		auto row_data = cereal::binary_data(mat.ptr(i), row_size);
+		ar & row_data;
+	}*/
+};
+
+
 /**
  * @brief Serialisation of a cv::Vec2f using cereal.
  *
@@ -111,6 +220,6 @@ void serialize(Archive& ar, cv::Vec2f& vec)
 	ar(vec[0], vec[1]);
 };
 
-}
+} /* namespace cv */
 
 #endif /* MATCEREALISATION_HPP_ */