Mahdi Naddaf

Mahdi Naddaf

Research Scientist

EnRisk

Biography

I am Mahdi Naddaf. Currently I am working as a research scientist at EnRisk. I was a research scientist at Ford Motor Company, Greenfield Labs, located in Palo Alto, California, United States. I was previously a postdoc at Galban Lab, University of Michigan. I completed my Doctoral of Engineering in Electrical Engineering with a focus on Computer Vision (Deep Learning) and Robotics at Lamar University in 2020 under the supervision of Dr. Hassan Zargarzadeh. I got my M.Sc. in Artificial Intelligence at the University of Southampton, the U.K, and a B.Sc. in Electrical Engineering- Control. You are most welcome to check out my research and recent publications.

Interests
  • Artificial Intelligence
  • Machine Vision - Deep Learning
  • Robotics
Education
  • Post-Doctoral research fellow, 2021

    University of Michigan, US

  • Doctoral degree in Electrical Engineering, 2020

    Lamar University, US

  • M.Sc. in Artificial Intelligence, 2012

    Southampton University, UK

Projects

*
Pick and Place using 3D deep object pose estimation with YuMi Cobot
  • Performing pick and place tasks with YuMi cobot with vision-based perception in the loop
  • Trained the deep neural network with fully synthetically generated data and labels using Isaac Sim
  • Utilizing ROS and customized drivers for controlling the YuMi including the end effectors
  • Extended to industrial robot platform ABB IRB-4600
  • Using domain randomization to robustify the vision algorithm against edge cases including change of lighting, texture, color and etc.
Pick and Place using 3D deep object pose estimation with YuMi Cobot
Motion Capture (MOCAP) System and Crazyflie Drones
  • Setting up Motion Analysis MOCAP system for RICS Lab.
  • Implemented a VRPN for utilizing swarm robots like crazyflie drones
  • The Setup is used by other graduate students for swarm robotics experiments
Motion Capture (MOCAP) System and Crazyflie Drones
Deep Learning Based Weld Defect Detection in X-ray Images
  • Developed Deep Learning based defect detector with Bayesian optimization using TensorFlow and Keras
  • Created database from scratch and pre-processing images using OpenCV, scikit learn, etc.
  • Funded by Stanley Black and Decker
  • Co-PI of the proposal
Deep Learning Based Weld Defect Detection in X-ray Images
Real-Time Road Crack Detection and Mapping
  • Developed and optimized deep learning based classification for crack classification
  • Implemented a heuristic mapping algorithm for mapping classified cracks
  • Implemented on both vehicle and drone platforms
  • Funded by Lamar University CICE
Real-Time Road Crack Detection and Mapping
Deep Learning Based Aquatic Robot for the Lionfish Remediation
  • Implemented Deep Learning based object detector on open source robot OpenROV to detect invasive species
  • Funded by Lamar University CAWAQ
  • Media coverage - Daily Planet Meet the Lionfish Hunter
Deep Learning Based Aquatic Robot for the Lionfish Remediation
3-PSP Parallel Robot with Fuzzy Force Controller
  • Designed and implemented a complete electrical interface for 3-PSP Parallel Robot, which is submitted as my B.Sc. final Project
  • Designed and implemented a Fuzzy Controller for 3-PSP Parallel Robot
3-PSP Parallel Robot with Fuzzy Force Controller

Slected Publications

(2023). Deep Bayesian-Assisted Keypoint Detection for Pose Estimation in Assembly Automation. Sensors.

PDF Cite DOI

(2022). Real-Time Explainable Multiclass Object Detection for Quality Assessment in 2-Dimensional Radiography Images. Complexity.

PDF Cite DOI

(2021). Defect detection and classification in welding using deep learning and digital radiography. Academic Press.

PDF Cite DOI

(2020). An efficient and scalable deep learning approach for road damage detection. 2020 IEEE International Conference on Big Data (Big Data).

PDF Cite Code DOI

(2019). Real-time road crack mapping using an optimized convolutional neural network. Complexity.

PDF Cite Video DOI

Gallery

Contact