# NgpSharp

**Repository Path**: GDIKodiak/ngp-sharp

## Basic Information

- **Project Name**: NgpSharp
- **Description**: a implementation of Instant-NGP with C# and CUDA Language.
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-09-30
- **Last Updated**: 2024-09-30

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# NGP Sharp

This is a implementation of Instant-NGP ([Instant-NGP](https://nvlabs.github.io/instant-ngp/assets/mueller2022instant.pdf))
with C# and CUDA Language.

project reference [HashNERF](https://github.com/johnny-walker/HashNeRF) and [toch-ngp](https://github.com/ashawkey/torch-ngp)

# Project dependence
Visual Studio 2022 or Visual Studio Code </br>
.NET 8.0 or heighter </br>
CUDA 12.1 or heighter </br>
libtorch 2.4 release or heighter </br>
TorchSharp-cuda-windows 0.103.0 or heighter </br>
NVIDIA RTX 3070 or heighter</br>

# Build
Just into NgpSharp folder and run
~~~
dotnet build
~~~
then wait building finish

# About TrainConsole
You could use NgpTrainConsole to train ngp model with NeRF blender dataset (like fox or lego dataset), into Program.cs and change datasetPath to your dataset path like:
~~~
        public const string DatasetPath = "E:\\(yourdataset)\\lego\\transforms_train.json";
        public const string EvalPath = "E:\\(yourdataset)\\lego\\transforms_val.json";
~~~
then run console and wait trainning finish.</br>
finaly in the outputpath:
~~~
public static string OutputPath = Path.Combine(AppContext.BaseDirectory, "output");
~~~
you will found eval output image

# TODO
GUI Supported </br>
LIFF Dataset Supported