Skip to content

Latest commit

 

History

History
106 lines (60 loc) · 4.27 KB

Synthetic-Data-UE4-DOPE.md

File metadata and controls

106 lines (60 loc) · 4.27 KB
Raw

Synthetic-Data-UE4-DOPE

How to use the Blender & UE4

STEP 1 - Download the NDDS Documentation

NVIDIA Deep learning Dataset Synthesizer (NDDS) Documentation

wget https://github.com/yehengchen/Synthetic-Data-UE4/blob/master/NDDS.pdf

STEP 2 - Installing NDDS

Windows Specific:

First, install Visual Studio. a. If you are installing Visual Studio or Visual Studio Community, use the 2017 version located here. (Other versions may not be compatible.)

i. We successfully tested NDDS using Visual Studio 2017 version 15.9.

b. When installing Visual Studio, you will need to perform a CUSTOM INSTALL. Make sure the following are selected

  • Desktop development with C++
  • Latest Windows 10 SDK
  • Windows 8.1 SDK (if you are still on Windows 8.1)

Download the following:

a. Epic's Launcher so that you can install Unreal Engine.

i. Due to Unreal's requirements, please always ensure that you create your directory starting with a letter and not a number.

b. Installation instructions are found here.

  1. Let's start with Unreal. Once you have downloaded Epic Launcher, install it and create an Epic account.

  2. Once installed start the Launcher, login with your account and go to the Unreal Engine tab on the upper left corner of the Launcher. From here, make sur eyou have Library highlighted and then click the " + " icon.

a. A pull down menu will appear with various engine versions. Please select Version 4.21.0

  1. With the engine is installed, start the Editor by pressing the Launch button. You will need to start the Editor at least once so that it sets file associations. Creating a blank default project is good enough.

  2. Close the Editor and Launcher for now.

Linux Specific:

a. for Unreal Engine, refer to the UE4 website instructions: https://wiki.unrealengine.com/Building_On_Linux Note NDDS uses UE version 4.21.0, with respect to the unreal first time setup documentation, use git clone -b 4.21

See also https://docs.unrealengine.com/en-US/Platforms/Linux/GettingStarted

b. Note that NDDS currently supports only Ubuntu 16.04.4 LTS

c. We successfully tested NDDS using NVIDIA driver versions 387.34 and 390.67

d. If you get this warning, select "More options", then "Skip conversion"

Installation

  1. Download NDDS:

a. Ensure LFS is installed: https://help.github.com/articles/installing-git-large-file-storage/.

Install the NDDS files for the dataset synthesizer.

git clone https://github.com/NVIDIA/Dataset_Synthesizer.git

  1. Navigate to the directory containing the files and find NDDS.uproject. This should be under the \Source sub-directory.

  2. Run NDDS.uproject and select "Yes" when it prompts you to rebuild binaries. The compilation will occur behind the scene and open the project wh en completed.

Note: Nvidia Deep Learning Dataset Synthesizer plugins may only be used within a project (game) -- hosting as engine plugins not yet supported.

STEP 3 - Installing Blender

Download [Blender]

Make a 3D model in solidworks or CAD etc.,then import a 3D model in Blender (.stl .dae ...) and export it as a fbx file then import that fbx file into UE4, fbx file is work in UE4.

STEP 4 - Run the NDDS

Open the Unreal Editor with the Dataset_Synthesizer/Source/NDDS.uproject, a default level called TestCapturer will load as indicated at the top left hand corner of the 3D view port. This level has a sample scene with a basic simulation capture set up.

STEP 5 - Train the DOPE model

Training code is also provided but not supported - train.py

python train.py --data ./cube/ --outf cube_1214  --gpuids 0 1 --epochs 120 --loginterval 1 --batchsize 32