Games for Machines repository

This repository is intended for useful resources and relevant research papers in the area of sim2real, synthetic dataset creation for image recognition and other CV tasks, game engine tools for reinforcement learning.

---

RESEARCH

Training with Synthetic datasets

Training Deep Networks with Synthetic Data: Bridging the Reality Gap by Domain Randomization (CVPR 2018)

[arXiv] [info] [video]

Deep Object Pose Estimation for Semantic Robotic Grasping of Household Objects (Conference on Robot Learning 2018)

[arXiv] [video] [info] [Code]

Sim2real

Virtual-to-real Deep Reinforcement Learning: Mobile Robots Mapless Navigation

[arXiv] [video]

VR-Goggles for Robots: Real-to-sim Domain Adaptation for Visual Control

[arXiv] [video]

Reinforcement learning environments and simulations tools

Unity

Unity is a general purpose real time game engine, with build-in graphics and physics engine as well as some specific integrations and tools for sim2real work or reinforcement learning.

Unity SimViz

[Info] [Video]

The Unity SimViz Solution Template package includes sensor models, referred to as SynCity Lite, scene models, and supporting documentation. The parts can work together or separately, depending on the use case.

*currently SimViz is in a closed beta stage

Sensor models included:

• RGB-D Camera
• LiDAR
• Radar
• IMU
• GPS

Unity ML-agents

[Info] [Code] [Video]

Public* Datasets (synthesized from 3D data)

*please check license for each dataset

ShapeNET [Link]

ObjectNet3D [Link]

ABC: A Big CAD Model Dataset For Geometric Deep Learning [arXiv]