NNEngine

A plugin for Unreal Engine that provides easy-to-use functions for accelerated machine learning inference callable from BP and C++ using ONNX Runtime native library

Product Details

Overview

Easy to use

By simply calling a few Blueprint nodes, you can load and run ONNX-format AI.
No C++ coding required. (Of course, it also supports execution from C++.)

Designed for best performance

Automatically optimizes the model when loaded by removing redundant nodes and fusing nodes. Supports GPU accelerations by DirectML, CUDA, and TensoRT.

With practical examples

Examples of human pose estimation, facial capture, and eye tracking using a single RGB camera are provided.

Features

・Loads ONNX models at runtime. Automatically optimizes the model when loaded.
・Runs ONNX models.
・(On Windows) Supports hardware acceleration by DirectML on DirectX 12 capable GPUs.
・(On Windows and Ubuntu) Supports hardware acceleration by CUDA and TensorRT on supported NVIDIA GPUs.
・(On Windows) Gets a list of GPU names available on the system.
・Processes (resize, crop, rotate) UTexture and converts it to int8 array.

※To use GPU acceleration with DirectML, a DirectX 12 capable GPU is required.
※To use GPU acceleration with CUDA and TensorRT, a supported NVIDIA GPU is required and the following versions of CUDA, cuDNN, and TensorRT are required to be installed.

Download

Fab↗

Manual

Manual↗

Video tutorials

Demo projects

NEngine plugin is required to execute demo.

Monocular Depth Estimation (GitHub)↗

Arbitrary Artistic Style Transfer (GitHub)↗

U2-Net Portrait-like image generation (GitHub)↗

Human pose estimation + Facial capture + Eye tracking (GitHub)↗

FAQ

. Does NNEngine plugin have a feature to estimate 3D human pose, like MocapForAll?

A. No.
 This plugin provides utilities to run AI.
 Demo project provides examples including 2D human pose estimation. But 3D pose estimation is out of scope of this product.