Add serialization code.

Source/UnrealCV/Private/Serialization.cpp

@@ -0,0 +1,90 @@
+#include "UnrealCVPrivate.h"
+#include "Serialization.h"
+#include "ImageWrapper.h"
+#include "cnpy.h"
+
+TArray<uint8> SerializationUtils::Array2Npy(const TArray<FFloat16Color>& ImageData, int32 Width, int32 Height, int32 Channel)
+{
+	float *TypePointer = nullptr; // Only used for determing the type
+
+	std::vector<int> Shape;
+	Shape.push_back(Height);
+	Shape.push_back(Width);
+	if (Channel != 1) Shape.push_back(Channel);
+
+	std::vector<char> NpyHeader = cnpy::create_npy_header(TypePointer, Shape);
+
+	// Append the actual data
+	// FIXME: A slow implementation to convert TArray<FFloat16Color> to binary.
+	/* A small size test
+	std::vector<char> NpyData;
+	for (int i = 0; i < 3 * 3 * 3; i++)
+	{
+		NpyData.push_back(i);
+	}
+	*/
+	// std::vector<char> NpyData;
+	std::vector<float> FloatData;
+	float DebugMin = 10e10, DebugMax = 0.0;
+
+	for (int i = 0; i < ImageData.Num(); i++)
+	{
+		if (Channel == 1)
+		{
+			float v = ImageData[i].R;
+			FloatData.push_back(ImageData[i].R);
+			// debug: Check the range of data
+			if (v < DebugMin) DebugMin = v;
+			if (v > DebugMax) DebugMax = v;
+		}
+		if (Channel == 3)
+		{
+			FloatData.push_back(ImageData[i].R);
+			FloatData.push_back(ImageData[i].G);
+			FloatData.push_back(ImageData[i].B); // TODO: Is this a correct order in numpy?
+		}
+	}
+	check(FloatData.size() == Width * Height * Channel);
+	// Convert to binary array
+	const unsigned char* bytes = reinterpret_cast<const unsigned char*>(&FloatData[0]);
+
+	// https://stackoverflow.com/questions/22629728/what-is-the-difference-between-char-and-unsigned-char
+	// https://stackoverflow.com/questions/11022099/convert-float-vector-to-byte-vector-and-back
+	std::vector<unsigned char> NpyData(bytes, bytes + sizeof(float) * FloatData.size());
+
+	NpyHeader.insert(NpyHeader.end(), NpyData.begin(), NpyData.end());
+
+	// FIXME: Find a more efficient implementation
+	TArray<uint8> BinaryData;
+	for (char Element : NpyHeader)
+	{
+		BinaryData.Add(Element);
+	}
+	return BinaryData;
+}
+
+TArray<uint8> SerializationUtils::Image2Png(const TArray<FColor>& Image, int Width, int Height)
+{
+	if (Image.Num() == 0 || Image.Num() != Width * Height)
+	{
+		return TArray<uint8>();
+	}
+	static IImageWrapperModule& ImageWrapperModule = FModuleManager::LoadModuleChecked<IImageWrapperModule>(FName("ImageWrapper"));
+	static IImageWrapperPtr ImageWrapper = ImageWrapperModule.CreateImageWrapper(EImageFormat::PNG);
+	ImageWrapper->SetRaw(Image.GetData(), Image.GetAllocatedSize(), Width, Height, ERGBFormat::BGRA, 8);
+	const TArray<uint8>& ImgData = ImageWrapper->GetCompressed(ImageCompression::Uncompressed);
+	return ImgData;
+}
+
+TArray<uint8> SerializationUtils::Image2Exr(const TArray<FFloat16Color>& FloatImage, int Width, int Height)
+{
+	if (FloatImage.Num() == 0 || FloatImage.Num() != Width * Height)
+	{
+		return TArray<uint8>();
+	}
+	static IImageWrapperModule& ImageWrapperModule = FModuleManager::LoadModuleChecked<IImageWrapperModule>(FName("ImageWrapper"));
+	static IImageWrapperPtr ImageWrapper = ImageWrapperModule.CreateImageWrapper(EImageFormat::EXR);
+	ImageWrapper->SetRaw(FloatImage.GetData(), FloatImage.GetAllocatedSize(), Width, Height, ERGBFormat::RGBA, 16);
+	const TArray<uint8>& ExrData = ImageWrapper->GetCompressed(ImageCompression::Uncompressed);
+	return ExrData;
+}

Source/UnrealCV/Public/Serialization.h

@@ -0,0 +1,9 @@
+#pragma once
+
+class SerializationUtils
+{
+public:
+	static TArray<uint8> Array2Npy(const TArray<FFloat16Color>& ImageData, int32 Width, int32 Height, int32 Channel);
+	static TArray<uint8> Image2Png(const TArray<FColor>& Image, int32 Width, int32 Height);
+	static TArray<uint8> Image2Exr(const TArray<FFloat16Color>& FloatImage, int Width, int Height);
+};