UnrealCV - CVPR 17 Demo

Welcome to the CVPR17 demo page of UnrealCV. This page contains what you saw in our demo station. Our demo time is 07/22 Morning 10:30 - 12:30 and Afternoon 15:00 - 17:00. Please share this link with your friends if you like our demo.

Our demo includes two parts, pre-recorded videos and live-demo

Video

The first video demonstrates various applications of UnrealCV, including:

  • Object mask generation in a virtual supermarket,
  • Bounding box generation and running a detection algorithm in a virtual scene,
  • Controlling hazardous factors to diagnose the robustness of stereo depth estimation algorithms,
  • Indoor navigation and object interaction.

The second video includes various 3D scenes that can be used in UnrealCV, including:

  • A static virtual scene, RealisticRendering,
  • A dynamic virtual scene, Matinee,
  • A virtual supermarket in which grocery items are randomly placed for training,
  • More virtual scenes that have been used in the research

Live demo

  • Faster-RCNN detection demo

    Move a virtual camera in a virtual room, see how well the detection algorithm works. Also demonstrate the ground truth that can be produced by a virtual scene using UnrealCV.

    You can reproduce this demo with two simple steps

    1. Download the RealisticRendering virtual scene from our model zoo http://unrealcv.org/model_zoo
    2. Follow the tutorial in this page http://docs.unrealcv.org/en/master/tutorials/faster_rcnn.html

  • Virtual supermarket demo

  • A virtual toy-robotics arm

Poster

The digital poster can be downloaded from here.

These papers use UnrealCV as a tool in the research

  • Qiao, Siyuan, Wei Shen, Weichao Qiu, Chenxi Liu, and Alan Yuille. “ScaleNet: Guiding Object Proposal Generation in Supermarkets and Beyond.” ICCV17
  • Roberts, Mike, Anh Truong, Debadeepta Dey, Sudipta Sinha, Ashish Kapoor, Neel Joshi, and Pat Hanrahan. “Submodular Trajectory Optimization for Aerial 3D Scanning.” ICCV17
  • Zhang, Yi, Weichao Qiu, Qi Chen, Xiaolin Hu, and Alan Yuille. “Unrealstereo: A synthetic dataset for analyzing stereo vision.” arXiv preprint arXiv:1612.04647 (2016).

More description about UnrealCV can be found in

Qiu, Weichao, and Alan Yuille. “UnrealCV: Connecting Computer Vision to Unreal Engine.” ECCV 2016 VARVAI workshop

See more papers and projects which utilize synthetic images here. https://github.com/unrealcv/synthetic-computer-vision

Notice

The license of UE4 projects used in the demo belongs to the content creator, please see the model zoo to see where you can get these UE4 projects.

UnrealCV is an open source project created by students of Johns Hopkins University and Peking University. The development of it is under the supervision of Prof. Alan Yuille and Prof. Yizhou Wang.

If anything of this page is unclear, please contact Weichao Qiu (qiuwch@gmail.com).