Adarash Kumar

PhD Researcher — AI for Autonomous Vehicle Testing

Professional Summary

Adarash Kumar is a final-year PhD researcher working on AI and computer vision for self-driving car testing. His research focuses on adversarial robustness of object detection systems in autonomous driving, using simulation environments and neural rendering to evaluate and improve safety-critical perception models.
Download CV

Education

PhD — AI for Self-Driving Car Testing

2022-09-01

Interests

Adversarial Machine Learning Computer Vision Autonomous Vehicle Safety Object Detection Neural Rendering Simulation-based Testing
🔬 My Research

My research investigates the adversarial robustness of object detection systems used in autonomous vehicles. I use physically-realisable adversarial attacks — camouflage patterns that look like natural environmental debris (snow, mud, dirt) — to fool state-of-the-art detectors while remaining visually inconspicuous.

Key contributions:

  • Photo-realistic neural renderer for generating adversarial vehicle textures in simulation
  • Adversarial camouflage that transfers across multiple detector architectures (EfficientDet, YOLOv5, SSD, Faster R-CNN)
  • Naturalistic constraint framework ensuring attack patterns resemble real-world environmental effects

I work with the CARLA driving simulator and PyTorch, combining neural rendering with adversarial optimisation to bridge the sim-to-real gap.

Featured Publications
Adversarial ML

Adversarial Vehicle Camouflage via Photo-Realistic Neural Rendering

We present a method for generating adversarial vehicle camouflage that fools object detectors while maintaining photo-realistic appearance. Using a neural renderer trained on the …

Adarash Kumar
•
Read more
Recent Publications
Adarash Kumar (2026). Adversarial Vehicle Camouflage via Photo-Realistic Neural Rendering. Submitted to IROS 2026.