---
title: Generate Synthetic Data of Dynamic Environments With a Custom Isaac Sim-Based Framework [S51570]
tags: GTC2023
description: Nvidia 2023 GTC session
---

# Generate Synthetic Data of Dynamic Environments With a Custom Isaac Sim-Based Framework [S51570]

# Nvidia issac sim
## Grade
GRADE is a system I developed to seamlessly manage the Isaac Sim simulation software to Generate Realistic Animated Dynamic Environments for Robotics Research.


### grade main componets
1. Readlistic Environmnets
2. Animaton Assets
3. Custom robots
4. A Customizable pipeline
5. Additional processing tools

### grade software pipeline
1. create environment
* Blender
* Unreal Engine
* Digital twin
2. Convert to USD format
3. Load it into Isaac Sim
4. Using sdk to set environment and camera to generate dataset

#### Verify
yolo people detection:
1. Train on synthetic data -> fine tune on S-coco dataset and full coco dataset
2. Train on S-coco dataset and full coco dataset

**Result: 1. is more accurate than 2.**

---
## 參考
* [https://github.com/eliabntt/GRADE-RR](https://github.com/eliabntt/GRADE-RR)

## Thank you! :dash:
You can find me on
- GitHub: https://github.com/shaung08
- Email: [e-mail]