James Lee

Javad Amirian

PhD Candidate in Robotics and AI

Rainbow Team, Inria Rennes

Contact Me

About Me

I am a PhD candidate at Inria Center in Rennes, and a member of Rainbow team which is specialized on Robotics, Computer Vision and Crowd Simulation. My PhD (supervised by Julien Pettre), is a part of CrowdBot project, a H2020 European project, aiming to develop platforms and algorithms to ensure safe navigation of social robots.
My main task is to develop tools for "Trajectory Prediction" of the agents around the robot, in order to improve the robot navigation in crowded environments. For that, I'm working with Crowd Simulation algorithms and developing Deep Learning models. We hope that this work, will also help to handicap people, by improving the performance of Semi-Autonomous Wheelchairs.
Before joining Inria, I was studying AI in Sharif University of Technology in Iran, followed by working in start-up companies in Tehran. You can find my CV here.
I'm also intrested in travelling and photography. You can check out some of my photos on Flicker, Unsplash, or Instagram. I'm also writing / translating about AI and Technology. Here you can find some of my posts (in Persian).

My Projects

CrowdBot H2020 project

Crowdbot is about “Safe Navigation of Robots in Dense Human Crowds”. The team is comprised of 7 European partners: Inria and Soft Bank Robotics (FR), ETHZ & EPFL (SW), RWTH & Locomotech (GR) and UCL (UK). The name “CROWDBOT” is an amalgam of human “crowd” and machine “bot.”

Recent Publications

OpenTraj: Assessing Prediction Complexity in Human Trajectories Datasets

J. Amirian, B. Zhang, F. Valente, J. José Baldelomar, J. Hayet, J. Pettré
ACCV 2020

We address the question of evaluating how complex is a given Human Trajectory Prediction dataset with respect to the prediction problem. For assessing a dataset complexity, we define a series of indicators around three concepts: Trajectory predictability; Trajectory regularity; Context complexity.

pdf   Code

Generalized Microscopic Crowd Simulation using Costs in Velocity Space (UMANS)

W. van Toll, F. Grzeskowiak, A. López, J. Amirian, F. Berton, J. Bruneau, B. Cabrero Daniel, A. Jovane, J. Pettré
I3D 2020

We present a novel framework that describes local agent navigation generically as optimizing a cost function in a velocity space. We show that many state-of-the-art algorithms can be translated to this framework, by combining a particular cost function with a particular optimization method.

pdf   Code

Social ways: Learning multi-modal distributions of pedestrian trajectories with GANs

Javad Amirian, Jean-Bernard Hayet, Julien Pettré
CVPR Workshop 2019

A novel approach for predicting the motion of pedestrians interacting with others. It uses a Generative Adversarial Network (GAN) to sample plausible predictions for any agent in the scene. As GANs are very susceptible to mode collapsing and dropping, we show that the recently proposed Info-GAN allows dramatic improvements in multi-modal pedestrian trajectory prediction to avoid these issues.

pdf   Code

Data-Driven Crowd Simulation with Generative Adversarial Networks

Javad Amirian, Wouter Van Toll, Jean-Bernard Hayet, Julien Pettré
CASA 2019

A novel data-driven crowd simulation method that can mimic the observed traffic of pedestrians in a given environment. Given a set of observed trajectories. We use Generative Adversarial Networks (GANs), to learn the properties of this set and generate new trajectories with similar properties.

pdf   Code

➡ To see the list of all of my publication, you can visit my Google scholar profile.

My Previous Projects

Online Camera Calibration [Sepehr Startup]

Sepehr is a computer vision startup that develop software solutions to sport teams to analyze their team performance. The company builds a video editing tool to help users create sport video clips, visualizing the player traces, adding graphical components to the video and so on. My task was to develop image processing and computer vision algorithms in order to provide high-level APIs for UI programmers. My main contribution was to design a system for estimating camera calibration parameters with corresponding lines of the court. This algorithm contains image filters to detect the lines, strategies for taking samples from the image, and a LM optimizer to find the calibration parameters.

Cyrus (Team of Small-size Soccer Robots) Open Source

I was co-founder and the leader of the student robotics team, Cyrus. The team was active since 2010 to 2016. I was also contributing to develop the Navigation of the robots, and the motion planning system.


Related publications

  • CYRUS 2016 Team Description Paper Javad Amiryan, Sina Raeessi, Pouya Payandeh, Bardia Nadimi, Navid Nouri, Mohamad Reza Kamali, Eslam Nazemi. Robocup 2016. [pdf] [Source Code]
  • Adaptive motion planning with artificial potential fields using a prior path Javad Amiryan, Mansour Jamzad. 3rd RSI International Conference on Robotics and Mechatronics (ICROM) 2015. [pdf]
  • Improvement of robot navigation using fuzzy method Mostafa Nazari, Javad Amiryan, Eslam Nazemi. 3rd Joint Conference of AI & Robotics and 5th RoboCup Iran Open International Symposium 2013. [pdf]

    • My GitHub

    You can follow my GitHub activities from here.

    GitHub Contributions Chart