I recently finished my M.Sc. in Computer Science at Mila/University of Montreal/Robotics and Embodied AI Lab (REAL) under supervision of Prof. Liam Paull. I have also been fortunate to collaborate with Dr. Anqi Xu on my current and past research works. My research focus was to develop learning framework for solving semantic visual navigation task. Most recently, I just finished working on Lifelong Topological Visual Navigation. Prior to this, I spent considerable time researching the topic of adversarial attacks and defenses, where I published PAT (ICCV’19), MJSMA (MAIS’18), and a review paper (pre-print). I was also lucky to have been concretely introduced to the research world by working with Prof. Jeff Orchard in the Neurocognitive Computing Lab during my undergrad days. In general, I love working on various aspects on building robots, whether it is hardware or software - whatever makes a great robot great. I also enjoy presenting my work, so please feel free to let me know if you want to learn more about it!
CV / Blog / Paper summaries
When I am not doing research, I like to play music (mainly acoustic and electric guitars), make educational videos, read books, writing blog posts or paper summaries, and go for a hike. I also recently started to explore the creative side of 3D modelling and animation, here is my coffee + donut:
Note: the coffee + donut were created from scratch, and the textures of the wall + table + ground plane are from Poliigon.
Please see my Google Scholar page for the complete list.
Lifelong Topological Visual Navigation
Rey Reza Wiyatno, Anqi Xu, Liam Paull
Paper / Webpage / Blog
A learning-based topological visual navigation method with graph update
strategies that improves lifelong navigation performance over time.
Our model can also be finetuned and work well in the real world.
Adversarial Examples in Modern Machine Learning: A Review
Rey Reza Wiyatno, Anqi Xu, Ousmane Dia, Archy de Berker
Paper / Adversarial Attack Blog / Adversarial Defense Blog
A review of adversarial attack and defense techniques.
Physical Adversarial Textures That Fool Visual Object Tracking
Rey Reza Wiyatno, Anqi Xu
Paper / Video / Blog
Fooling object tracker in the real world with inconspicuous-looking textures.
Featured Applied Projects
Please see my CV for the complete list.
Duckietown: Lane Following with Other Vehicles and Obstacles
Course Project, AI Driving Olympics, 2019
Lane following and collision avoidance using light-weight semantic
segmentation model and pure pursuit controller.
Inspeksi: Intelligent Manufacturing Visual Inspection Tool
Winner of Autodesk Canada Capstone Design Award, 2019
A low-cost solution for automated visual inspection tool.
Autonomous Drone Tracking
Work Project, 2018
Robot: Parrot Bebop 2
Autonomous drone tracking using GOTURN and PID controller.
Trident ROV Payload
Work Project, 2016
Robot: Sofar Trident
Injection moldable external payload for underwater robot with a custom WiFi