diff --git a/CMakeLists.txt b/CMakeLists.txt
index 162568b..c61c7d4 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -73,6 +73,7 @@ set(headers
src/sceneStructs.h
src/preview.h
src/utilities.h
+ external/include/tiny_gltf.h
)
set(sources
diff --git a/README.md b/README.md
index f044c82..e67ec1b 100644
--- a/README.md
+++ b/README.md
@@ -3,11 +3,142 @@ CUDA Denoiser For CUDA Path Tracer
**University of Pennsylvania, CIS 565: GPU Programming and Architecture, Project 4**
-* (TODO) YOUR NAME HERE
-* Tested on: (TODO) Windows 22, i7-2222 @ 2.22GHz 22GB, GTX 222 222MB (Moore 2222 Lab)
+* Utkarsh Dwivedi
+ * [LinkedIn](https://www.linkedin.com/in/udwivedi/), [personal website](https://utkarshdwivedi.com/)
+* Tested on: Windows 11 Home, AMD Ryzen 7 5800H @ 3.2GHz 16 GB, Nvidia GeForce RTX 3060 Laptop GPU 6 GB
-### (TODO: Your README)
+## Introduction
-*DO NOT* leave the README to the last minute! It is a crucial part of the
-project, and we will not be able to grade you without a good README.
+|
|
|
+|:-:|:-:|
+|Original 100 samples per pixel|Denoised|
+This is a CUDA implementation of the paper ["Edge-Avoiding A-Trous Wavelet Transform for fast Global Illumination Filtering"](https://jo.dreggn.org/home/2010_atrous.pdf), which describes a fast-approximation for a Gaussian-blur based denoiser for noisy path traced images coupled with an edge-avouding solution. This denoiser helps significantly reduce render times by reducing the overall samples per pixel required to get a denoised render that looks comparable to a render that would require many more samples per pixel to look similar.
+
+This project is an extension of my [CUDA Path Tracer](https://github.com/utkarshdwivedi3997/CIS-565-Project3-CUDA-Path-Tracer).
+
+## Process
+
+### 1. Generate noisy path-traced result
+
+This can be something generated with very few samples per pixel (in my experience as low as 40 spp work well with lambertian diffuse surfaces).
+
+|
|
+|:-:|
+|100 samples per pixel|
+
+### 2. Generate G-Buffer
+
+The paper explains using a G-buffer with world-space positions and world-space normals as a texture. This project's CUDA implementation simply makes a buffer of these values in the first path tracing sample and sends them over to the GPU for use later in the denoising process.
+
+|
|
|
|
+|:-:|:-:|:-:|
+|World Space Positions|World Space Normals|Depth (only visualised, not needed in G-buffer)|
+
+### 3. Blur noisy pathtraced image using A trous wavelet transform
+
+I'm not a fan of fancy names. What the A trous wavelet transformation described in this paper actually does is a **fast approximation of the Gaussian blur applied over iteratively increasing filter sizes**.
+
+This works as follows: in Gaussian blur, we look at the neighbouring pixels of the current pixel and average out its colours based on a weight kernel that describes how much each pixel's colour contributes to the final average colour of the current pixel. "Denoising" here works by increasing this kernel size over multiple iterations, so we first look at a kernel of 5x5, then 7x7, then 9x9 and so on. This can become expensive, as we start to include more and more pixels for computation with increasing kernel sizes.
+
+This is where "a-trous" comes in. In very simple terms, this is still a Gaussian blur, but *with holes*. When kernel size is increased in successive iterations, the number of pixels sampled for averaging colour is kept constistent: every iteration, every 2i *th* pixel is sampled, where *i* is the iteration number. This keeps the kernel size constant, and the iterative blur algorithm fast.
+
+|
|
+|:-:|
+|3 iterations of the fast approximated Gaussian blur kernel (image source: see [paper](https://jo.dreggn.org/home/2010_atrous.pdf))|
+
+This blur looks like this:
+|
|
|
|
+|:-:|:-:|:-:|
+|Fast Approximated Blur 3 Iterations|Fast Approximated Blur 5 Iterations|Photoshop's Gaussian Blur|
+
+### 4. Apply edge-stopping function
+
+Finally, we apply an edge stopping function that uses the position, normal from the G-buffer and the colour values from the noisy pathtraced result to maintain sharp edges in the blurred output, based on user-defined weightage given to each of the parameters.
+
+||
|
+|:-:|:-:|
+|Original 100 samples per pixel|Denoised|
+
+## Performance and Quality Analysis
+
+### Denoise time compared to path-tracing time
+
+This comparison was done using a filter size of 65x65 (5 denoiser iterations) on the cornell box scene with a specular sphere, by varying the samples per pixel taken during path tracing.
+
+
+
+Since the denoising happens only once at the end of the path-tracing, it does not have a significant cost if used. Since at least 5 samples are required to get a good result anyway, we can use the denoiser without much worry about its performance cost, as that is definitely not the bottleneck.
+
+What is interesting, however, is that the samples per pixel required to produce an "acceptable" result can be significantly lowered by using this denoiser.
+
+|
||
+|:-:|:-:|
+|Path-traced (5000 samples per pixel)
Render time: ~1 minute|Path-traced (100 samples per pixel) + Denoised
Render time: 1.03 seconds|
+
+The above denoised image does not look significantly different from the path-traced image, but does cut the render cost by almost 60x!
+
+### Denoise time at different resolutions
+
+This comparison was done using a filter size of 65x65 (5 denoiser iterations) on the cornell box scene with a specular sphere, with increasing image resolution.
+
+
+
+The time to denoise images with exponentially growing scales also scales exponentially. With that said, when compared to the time taken to path-trace an acceptable image, the denoise time is insignificant, as the path-tracing time is what scales up exponentially as well.
+
+### Denoise time with varying filter sizes
+
+This comparison was done using varying filter sizes on the cornell box scene with a specular sphere, with a constant image resolution of 800x800 pixels and 100 samples per pixel.
+
+|
|
|
|
+|:-:|:-:|:-:|
+|5x5 fitler (1 denoise iteration)|9x9 filter (2 denoiser iterations)|17x17 filter (3 denoiser iterations)|
+|
|
|
|
+|33x33 fitler (4 denoise iteration)|65x65 filter (5 denoiser iterations)|129x129 filter (6 denoiser iterations)|
+
+About 5 denoiser iterations are enough to get an acceptable result. The "gain" in visual quality does not seem to scale linearly with increasing filter size, as the gains seem to become lesser and lesser. The difference between the denoised results from 5 and 6 denoiser iterations are not significant enough to justify using more denoiser iterations.
+
+
+
+The time taken to denoise with exponentially increasing filter sizes scales linearly only, since exponentially increasing filter sizes only scale the denoiser iterations linearly (by 1 with each power of 2 filter size). This means that the filter size is not too much of a concern and will not be the primary bottleneck in path-tracing.
+
+### Denoiser vs material types
+
+**Lambertian Diffuse**
+
+|
|
|
|
+|:-:|:-:|:-:|
+|Path-traced (5000 samples per pixel)
Render time: ~1 minute|Path-traced (100 samples per pixel) + Denoised
Render time: <1 second|Diff|
+
+In the diffuse case, the denoiser works best. The denoised results look acceptable, though the only problematic areas are around the edges of the light source. This is an unfortunate side effect of the way this denoiser works.
+
+**Perfect Specular**
+
+|||
|
+|:-:|:-:|:-:|
+|Path-traced (5000 samples per pixel)
Render time: ~1 minute|Path-traced (100 samples per pixel) + Denoised
Render time: 1.03 seconds|Diff|
+
+While the result does not look that bad on the specular sphere, it is smoothed out a bit too much, as the normals on the sphere are changing constantly and its difficult to find any edge here for the edge stopping.
+
+**Refractive Materials**
+
+|
|
|
|
+|:-:|:-:|:-:|
+|Path-traced (10000 samples per pixel)
Render time: ~3.2 minutes|Path-traced (100 samples per pixel) + Denoised
Render time: 5 seconds|Diff|
+
+This is where this algorithm starts to fail. Most of the detail that is refracted through glass is blurred out too significantly to be able to be interpreted as refraction. In this example image, even the specular highlights on the icosphere behind the glass monkey are blurred out in the denoiser to the point where it looks like a simple diffuse icosphere.
+
+### Denoiser vs differently sized light sources
+
+||
|
+|:-:|:-:|
+|Path-traced (100 samples per pixel) + Denoised
Render time: 1.03 seconds|Path-traced (100 samples per pixel) + Denoised
Render time: 0.82 seconds|
+
+This result is expected. With a larger light source, there would be less noise in the path-traced result simply because a lot more rays would reach the light source. As a result, denoising a less noisy image would result in a smoother looking render.
+
+## References
+
+- Adam Mally's CIS 561 Advanced Computer Graphics course at University of Pennsylvania
+- TinyGLTF
+- PBRT
+- [Edge-Avoiding A-Trous Wavelet Transform for fast Global Illumination Filtering](https://jo.dreggn.org/home/2010_atrous.pdf)
\ No newline at end of file
diff --git a/cmake/CUDAComputesList.cmake b/cmake/CUDAComputesList.cmake
index 87629ae..3156a63 100644
--- a/cmake/CUDAComputesList.cmake
+++ b/cmake/CUDAComputesList.cmake
@@ -60,6 +60,8 @@ IF( CUDA_COMPUTE_20
OR CUDA_COMPUTE_70
OR CUDA_COMPUTE_72
OR CUDA_COMPUTE_75
+ OR CUDA_COMPUTE_80
+ OR CUDA_COMPUTE_86
)
SET(FALLBACK OFF)
ELSE()
@@ -70,8 +72,8 @@ LIST(LENGTH COMPUTES_DETECTED_LIST COMPUTES_LEN)
IF(${COMPUTES_LEN} EQUAL 0 AND ${FALLBACK})
MESSAGE(STATUS "You can use -DCOMPUTES_DETECTED_LIST=\"AB;XY\" (semicolon separated list of CUDA Compute versions to enable the specified computes")
MESSAGE(STATUS "Individual compute versions flags are also available under CMake Advance options")
- LIST(APPEND COMPUTES_DETECTED_LIST "30" "50" "60" "70")
- MESSAGE(STATUS "No computes detected. Fall back to 30, 50, 60 70")
+ LIST(APPEND COMPUTES_DETECTED_LIST "30" "50" "60" "70" "80")
+ MESSAGE(STATUS "No computes detected. Fall back to 30, 50, 60, 70, 80")
ENDIF()
LIST(LENGTH COMPUTES_DETECTED_LIST COMPUTES_LEN)
@@ -90,7 +92,7 @@ MACRO(SET_COMPUTE VERSION)
ENDMACRO(SET_COMPUTE)
# Iterate over compute versions. Create variables and enable computes if needed
-FOREACH(VER 20 30 32 35 37 50 52 53 60 61 62 70 72 75)
+FOREACH(VER 20 30 32 35 37 50 52 53 60 61 62 70 72 75 80 86)
OPTION(CUDA_COMPUTE_${VER} "CUDA Compute Capability ${VER}" OFF)
MARK_AS_ADVANCED(CUDA_COMPUTE_${VER})
IF(${CUDA_COMPUTE_${VER}})
diff --git a/cmake/FindGLFW.cmake b/cmake/FindGLFW.cmake
index 014ae83..9cf3786 100644
--- a/cmake/FindGLFW.cmake
+++ b/cmake/FindGLFW.cmake
@@ -20,57 +20,57 @@
include(FindPackageHandleStandardArgs)
if (WIN32)
- # Find include files
- find_path(
- GLFW_INCLUDE_DIR
- NAMES GLFW/glfw3.h
- PATHS
- $ENV{PROGRAMFILES}/include
- ${GLFW_ROOT_DIR}/include
- DOC "The directory where GLFW/glfw.h resides")
+ # Find include files
+ find_path(
+ GLFW_INCLUDE_DIR
+ NAMES GLFW/glfw3.h
+ PATHS
+ $ENV{PROGRAMFILES}/include
+ ${GLFW_ROOT_DIR}/include
+ DOC "The directory where GLFW/glfw.h resides")
- # Use glfw3.lib for static library
- if (GLFW_USE_STATIC_LIBS)
- set(GLFW_LIBRARY_NAME glfw3)
- else()
- set(GLFW_LIBRARY_NAME glfw3dll)
- endif()
+ # Use glfw3.lib for static library
+ if (GLFW_USE_STATIC_LIBS)
+ set(GLFW_LIBRARY_NAME glfw3)
+ else()
+ set(GLFW_LIBRARY_NAME glfw3dll)
+ endif()
- # Find library files
- find_library(
- GLFW_LIBRARY
- NAMES ${GLFW_LIBRARY_NAME}
- PATHS
- $ENV{PROGRAMFILES}/lib
- ${GLFW_ROOT_DIR}/lib)
+ # Find library files
+ find_library(
+ GLFW_LIBRARY
+ NAMES ${GLFW_LIBRARY_NAME}
+ PATHS
+ $ENV{PROGRAMFILES}/lib
+ ${GLFW_ROOT_DIR}/lib)
- unset(GLFW_LIBRARY_NAME)
+ unset(GLFW_LIBRARY_NAME)
else()
- # Find include files
- find_path(
- GLFW_INCLUDE_DIR
- NAMES GLFW/glfw.h
- PATHS
- /usr/include
- /usr/local/include
- /sw/include
- /opt/local/include
- DOC "The directory where GL/glfw.h resides")
+ # Find include files
+ find_path(
+ GLFW_INCLUDE_DIR
+ NAMES GLFW/glfw.h
+ PATHS
+ /usr/include
+ /usr/local/include
+ /sw/include
+ /opt/local/include
+ DOC "The directory where GL/glfw.h resides")
- # Find library files
- # Try to use static libraries
- find_library(
- GLFW_LIBRARY
- NAMES glfw3
- PATHS
- /usr/lib64
- /usr/lib
- /usr/local/lib64
- /usr/local/lib
- /sw/lib
- /opt/local/lib
- ${GLFW_ROOT_DIR}/lib
- DOC "The GLFW library")
+ # Find library files
+ # Try to use static libraries
+ find_library(
+ GLFW_LIBRARY
+ NAMES glfw3
+ PATHS
+ /usr/lib64
+ /usr/lib
+ /usr/local/lib64
+ /usr/local/lib
+ /sw/lib
+ /opt/local/lib
+ ${GLFW_ROOT_DIR}/lib
+ DOC "The GLFW library")
endif()
# Handle REQUIRD argument, define *_FOUND variable
@@ -78,8 +78,8 @@ find_package_handle_standard_args(GLFW DEFAULT_MSG GLFW_INCLUDE_DIR GLFW_LIBRARY
# Define GLFW_LIBRARIES and GLFW_INCLUDE_DIRS
if (GLFW_FOUND)
- set(GLFW_LIBRARIES ${OPENGL_LIBRARIES} ${GLFW_LIBRARY})
- set(GLFW_INCLUDE_DIRS ${GLFW_INCLUDE_DIR})
+ set(GLFW_LIBRARIES ${OPENGL_LIBRARIES} ${GLFW_LIBRARY})
+ set(GLFW_INCLUDE_DIRS ${GLFW_INCLUDE_DIR})
endif()
# Hide some variables
diff --git a/cmake/FindGLM.cmake b/cmake/FindGLM.cmake
index e2d45a8..55086f6 100644
--- a/cmake/FindGLM.cmake
+++ b/cmake/FindGLM.cmake
@@ -2,12 +2,12 @@
# Find GLM
#
# Try to find GLM : OpenGL Mathematics.
-# This module defines
+# This module defines
# - GLM_INCLUDE_DIRS
# - GLM_FOUND
#
# The following variables can be set as arguments for the module.
-# - GLM_ROOT_DIR : Root library directory of GLM
+# - GLM_ROOT_DIR : Root library directory of GLM
#
# References:
# - https://github.com/Groovounet/glm/blob/master/util/FindGLM.cmake
@@ -18,26 +18,26 @@
include(FindPackageHandleStandardArgs)
if (WIN32)
- # Find include files
- find_path(
- GLM_INCLUDE_DIR
- NAMES glm/glm.hpp
- PATHS
- $ENV{PROGRAMFILES}/include
- ${GLM_ROOT_DIR}/include
- DOC "The directory where glm/glm.hpp resides")
+ # Find include files
+ find_path(
+ GLM_INCLUDE_DIR
+ NAMES glm/glm.hpp
+ PATHS
+ $ENV{PROGRAMFILES}/include
+ ${GLM_ROOT_DIR}/include
+ DOC "The directory where glm/glm.hpp resides")
else()
- # Find include files
- find_path(
- GLM_INCLUDE_DIR
- NAMES glm/glm.hpp
- PATHS
- /usr/include
- /usr/local/include
- /sw/include
- /opt/local/include
- ${GLM_ROOT_DIR}/include
- DOC "The directory where glm/glm.hpp resides")
+ # Find include files
+ find_path(
+ GLM_INCLUDE_DIR
+ NAMES glm/glm.hpp
+ PATHS
+ /usr/include
+ /usr/local/include
+ /sw/include
+ /opt/local/include
+ ${GLM_ROOT_DIR}/include
+ DOC "The directory where glm/glm.hpp resides")
endif()
# Handle REQUIRD argument, define *_FOUND variable
@@ -45,7 +45,7 @@ find_package_handle_standard_args(GLM DEFAULT_MSG GLM_INCLUDE_DIR)
# Define GLM_INCLUDE_DIRS
if (GLM_FOUND)
- set(GLM_INCLUDE_DIRS ${GLM_INCLUDE_DIR})
+ set(GLM_INCLUDE_DIRS ${GLM_INCLUDE_DIR})
endif()
# Hide some variables
diff --git a/external/include/GL/Copying.txt b/external/include/GL/Copying.txt
deleted file mode 100644
index fc36ad9..0000000
--- a/external/include/GL/Copying.txt
+++ /dev/null
@@ -1,27 +0,0 @@
-
- Freeglut Copyright
- ------------------
-
- Freeglut code without an explicit copyright is covered by the following
- copyright:
-
- Copyright (c) 1999-2000 Pawel W. Olszta. All Rights Reserved.
- Permission is hereby granted, free of charge, to any person obtaining a copy
- of this software and associated documentation files (the "Software"), to deal
- in the Software without restriction, including without limitation the rights
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
- copies or substantial portions of the Software.
-
- The above copyright notice and this permission notice shall be included in
- all copies or substantial portions of the Software.
-
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- PAWEL W. OLSZTA BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
- IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
- Except as contained in this notice, the name of Pawel W. Olszta shall not be
- used in advertising or otherwise to promote the sale, use or other dealings
- in this Software without prior written authorization from Pawel W. Olszta.
diff --git a/external/include/GL/Readme.txt b/external/include/GL/Readme.txt
deleted file mode 100644
index f454169..0000000
--- a/external/include/GL/Readme.txt
+++ /dev/null
@@ -1,101 +0,0 @@
-freeglut 2.8.1-1.mp for MSVC
-
-This package contains freeglut import libraries, headers, and Windows DLLs.
-These allow 32 and 64 bit GLUT applications to be compiled on Windows using
-Microsoft Visual C++.
-
-For more information on freeglut, visit http://freeglut.sourceforge.net/.
-
-
-Installation
-
-Create a folder on your PC which is readable by all users, for example
-“C:\Program Files\Common Files\MSVC\freeglut\” on a typical Windows system. Copy
-the “lib\” and “include\” folders from this zip archive to that location.
-
-The appropriate freeglut DLL can either be placed in the same folder as your
-application, or can be installed in a system-wide folder which appears in your
-%PATH% environment variable. Be careful not to mix the 32 bit DLL up with the 64
-bit DLL, as they are not interchangeable.
-
-
-Compiling 32 bit Applications
-
-To create a 32 bit freeglut application, create a new Win32 C++ project in MSVC.
-From the “Win32 Application Wizard”, choose a “Windows application”, check the
-“Empty project” box, and submit.
-
-You’ll now need to configure the compiler and linker settings. Open up the
-project properties, and select “All Configurations” (this is necessary to ensure
-our changes are applied for both debug and release builds). Open up the
-“general” section under “C/C++”, and configure the “include\” folder you created
-above as an “Additional Include Directory”. If you have more than one GLUT
-package which contains a “glut.h” file, it’s important to ensure that the
-freeglut include folder appears above all other GLUT include folders.
-
-Now open up the “general” section under “Linker”, and configure the “lib\”
-folder you created above as an “Additional Library Directory”. A freeglut
-application depends on the import libraries “freeglut.lib” and “opengl32.lib”,
-which can be configured under the “Input” section. However, it shouldn’t be
-necessary to explicitly state these dependencies, since the freeglut headers
-handle this for you. Now open the “Advanced” section, and enter “mainCRTStartup”
-as the “Entry Point” for your application. This is necessary because GLUT
-applications use “main” as the application entry point, not “WinMain”—without it
-you’ll get an undefined reference when you try to link your application.
-
-That’s all of your project properties configured, so you can now add source
-files to your project and build the application. If you want your application to
-be compatible with GLUT, you should “#include ”. If you want to use
-freeglut specific extensions, you should “#include ” instead.
-
-Don’t forget to either include the freeglut DLL when distributing applications,
-or provide your users with some method of obtaining it if they don’t already
-have it!
-
-
-Compiling 64 bit Applications
-
-Building 64 bit applications is almost identical to building 32 bit applications.
-When you use the configuration manager to add the x64 platform, it’s easiest to
-copy the settings from the Win32 platform. If you do so, it’s then only necessary
-to change the “Additional Library Directories” configuration so that it
-references the directory containing the 64 bit import library rather
-than the 32 bit one.
-
-
-Problems?
-
-If you have problems using this package (compiler / linker errors etc.), please
-check that you have followed all of the steps in this readme file correctly.
-Almost all of the problems which are reported with these packages are due to
-missing a step or not doing it correctly, for example trying to build a 32 bit
-app against the 64 bit import library. If you have followed all of the steps
-correctly but your application still fails to build, try building a very simple
-but functional program (the example at
-http://www.transmissionzero.co.uk/computing/using-glut-with-mingw/ works fine
-with MSVC). A lot of people try to build very complex applications after
-installing these packages, and often the error is with the application code or
-other library dependencies rather than freeglut.
-
-If you still can’t get it working after trying to compile a simple application,
-then please get in touch via http://www.transmissionzero.co.uk/contact/,
-providing as much detail as you can. Please don’t complain to the freeglut guys
-unless you’re sure it’s a freeglut bug, and have reproduced the issue after
-compiling freeglut from the latest SVN version—if that’s still the case, I’m sure
-they would appreciate a bug report or a patch.
-
-
-Changelog
-
-2013–05–11: Release 2.8.1-1.mp
-
- • First 2.8.1 MSVC release. I’ve built the package using Visual Studio 2012,
- and the only change I’ve made is to the DLL version resource—I’ve changed
- the description so that my MinGW and MSVC builds are distinguishable from
- each other (and other builds) using Windows Explorer.
-
-
-Martin Payne
-2013–05–11
-
-http://www.transmissionzero.co.uk/
diff --git a/external/include/json.hpp b/external/include/json.hpp
new file mode 100644
index 0000000..87475ab
--- /dev/null
+++ b/external/include/json.hpp
@@ -0,0 +1,26753 @@
+/*
+ __ _____ _____ _____
+ __| | __| | | | JSON for Modern C++
+| | |__ | | | | | | version 3.10.4
+|_____|_____|_____|_|___| https://github.com/nlohmann/json
+
+Licensed under the MIT License .
+SPDX-License-Identifier: MIT
+Copyright (c) 2013-2019 Niels Lohmann .
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+*/
+
+#ifndef INCLUDE_NLOHMANN_JSON_HPP_
+#define INCLUDE_NLOHMANN_JSON_HPP_
+
+#define NLOHMANN_JSON_VERSION_MAJOR 3
+#define NLOHMANN_JSON_VERSION_MINOR 10
+#define NLOHMANN_JSON_VERSION_PATCH 4
+
+#include // all_of, find, for_each
+#include // nullptr_t, ptrdiff_t, size_t
+#include // hash, less
+#include // initializer_list
+#ifndef JSON_NO_IO
+ #include // istream, ostream
+#endif // JSON_NO_IO
+#include // random_access_iterator_tag
+#include // unique_ptr
+#include // accumulate
+#include // string, stoi, to_string
+#include // declval, forward, move, pair, swap
+#include // vector
+
+// #include
+
+
+#include
+#include
+
+// #include
+
+
+#include // transform
+#include // array
+#include // forward_list
+#include // inserter, front_inserter, end
+#include