Arsalan Mousavian

Address: NVIDIA
4545 Roosevelt Way, 6th Floor
Seattle, WA, 98105
Email: amousavi AT gmu DOT edu

2004 - 2010 2010 - 2013 June-Aug 2015 June-Aug 2016 June-Aug 2017 Aug 2017 - May 2018 2013 - 2018 2018 - now

Since August 2018, I am a research scientist at NVIDIA robotics research lab in Seattle. My research interests lies in using 3D perception for robotics tasks. These days, I am mainly working on model-free object manipulation in unconstrained environments and 3D understanding of the scene from RGB-D images. Prior to that, I graduated with a PhD from the Computer Science department at George Mason University where I was working under the supervision of Prof. Jana Kosecka. During my PhD, I worked on a variety of computer vision problems for robot perception such as: vision based navigation, object pose estimation, object detection, semantic segmentation and image based localization. I was fortunate to work with amazing collaborators in different research groups in industry. I did internships at Google Brain Robotics, Zoox, and Google StreetView during my PhD.



The Best of Both Modes: Separately Leveraging RGB and Depth for Unseen Object Instance Segmentation
Christopher Xie, Yu Xiang, Arsalan Mousavian, Dieter Fox
arxiv preprint, 2019
[Paper] [Video] [Projet Page] [bibtex]


6-DOF GraspNet: Variational Grasp Generation for Object Manipulation
Arsalan Mousavian, Clemens Eppner, Dieter Fox
International Conference on Computer Vision (ICCV), Seoul, South Korea, 2019
[Paper] [Video] [bibtex]


PoseRBPF: A Rao-Blackwellized Particle Filter for 6D Object Pose Tracking
Xinke Deng, Arsalan Mousavian, Yu Xiang, Fei Xia, Tim Bretl, Dieter Fox
Robotics: Science and Systems Conference (RSS) Freiburg, Germany, 2019
[Paper] [Video] [TechXplore Article] [bibtex]


Visual Represenatations for Semantic Target Driven Navigation
Arsalan Mousavian, Alexander Toshev, Marek Fiser, Jana Kosecka, James Davidson
International Conference on Robotics and Automation (ICRA) Montreal, Canada, 2019
[Paper] [bibtex]


Synthesizing Training Data for Object Detection in Indoor Scenes
Georgios Georgakis, Arsalan Mousavian, Alexander C. Berg, Jana Kosecka
Robotics: Science and Systems Conference (RSS), Cambridge MA, 2017
[Paper] [bibtex]


3D Bounding Box Estimation Using Deep Learning and Geometry
Arsalan Mousavian, Dragomir Anguelov, John Flynn, Jana Kosecka
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu HI, 2017
[Paper] [Supplementary Material] [bibtex] [IEEE Spectrum][Forbes]
Estimated 3D Boxes on the split of 3DVP: [Download Link]


Joint Semantic Segmentation and Depth Estimation with Deep Convolutional Networks
Arsalan Mousavian, Hamed Pirsiavash, Jana Kosecka
International Conference on 3DVision(3DV), Stanford CA, 2016
[Paper] [bibtex]


Multiview RGB-D Dataset for Object Instance Detection
Georgios Georgakis, Md Alimoor Reza, Arsalan Mousavian, Phi-Hung Le, Jana Kosecka
International Conference on 3DVision(3DV), Stanford CA, 2016
[Paper] [Dataset Link] [bibtex]


Semantic Image Based Geolocation Given a Map
Arsalan Mousavian, Jana Kosecka
arXiv 2016


Deep Convolutional Features for Image Based Retrieval and Scene Categorization
Arsalan Mousavian, Jana Kosecka
arXiv 2015
[Paper] [bibtex]


Semantically Guided Location Recognition for Outdoors Scenes
Arsalan Mousavian, Jana Kosecka, Jyh-Ming Lien.
International Conference on Automation and Robotics (ICRA), Seattle WA, 2015
[Paper] [bibtex]


Semantically Aware Bag-of-words for Localization
Arsalan Mousavian, Jana Kosecka.
Computer Vision and Pattern Recognition 2015 workshop on Semantic for Visual Reconstruction, Localization, and Mapping (CVPRW), Boston, MA, 2015
[Paper] [bibtex]