Dheeraj Khanna

I'm a

About Me

Computer Vision Engineer & Researcher passionate about solving real-world problems with AI

Welcome to my portfolio!

I'm Dheeraj, a Computer Vision Engineer and MASc graduate from the University of Waterloo, specializing in advanced deep learning and computer vision solutions.

🎯 Current Focus

• Multi-object tracking systems
• Motion prediction algorithms
• Automated annotation pipelines

🏆 Recent Achievements

• Papers accepted at CVPR Workshops 2025
• Published at CRV 2024
• 3+ years industry experience

My expertise spans from academic research in Multi-Object Tracking, sequence models and self-attention techniques to practical applications in autonomous vehicles, including sensor calibration, and edge device deployment. I'm passionate about applying AI and robotics to create meaningful impact in the real world.

Explore my projects, publications, and technical experiences throughout this portfolio. Feel free to connect - I'm always excited to discuss innovative AI solutions and collaborate on impactful projects!

Recent Updates

Latest accomplishments, milestones, and exciting developments

🎉 April 2025: Our paper "Attention-Mamba for Multi-Object Tracking" was accepted at the Conference on Robots and Vision (CRV) 2025.

🎉 April 2025: Our paper "SportMamba: Adaptive Non-Linear Multi-Object Tracking with State Space Models for Team Sports" was accepted at the CVPR Workshop 2025.

🎓 March 2025: Successfully submitted my thesis "Multi-Object Tracking using Mamba and an Investigation into Data Association Strategies" and graduated with MASc in Systems Design Engineering!

📄 April 2024: Our paper "POPCat: Propagation of particles for complex annotation tasks" was published at the Conference of Robots and Vision (CRV) 2024.

🔄 January 2023: Changed my program at the University of Waterloo to Master of Applied Science in System Design Engineering (MASc SYDE) under Prof. John Zelek.

💼 February 2022: Joined LENS AI as a Computer Vision Research Engineer.

📖 November 2020: Our work at IIITD, "Recipedb: A resource for exploring recipes", got published in Database Journal.

🏆 July 2020: 2nd runner-up in Microsoft Hackathon 2020. Watch demo.

Technical Skills

Core technologies and frameworks I work with regularly

Python 95%

PyTorch 90%

OpenCV 90%

C++ 80%

Docker 80%

Pandas 85%

Professional Experience

My journey through industry and research positions

Graduate Student Researcher | VIP Lab

September 2022 - Present

University of Waterloo, Ontario

Benchmarked multiple object tracking datasets with state-of-the-art techniques including Transformers, Diffusion models, Graph Neural Networks, Mamba, and motion models such as Kalman Filters
Created a novel automatic bounding box annotation pipeline to annotate videos with multiple objects achieving 10 FPS speed using Point Tracking (PIPs, TAPIR), Segment Anything (SAM), and Yolo-V8
Developed proprietary algorithms to calculate velocities of assembly parts in bowl feeder mechanisms, enhancing production yield and reducing manual intervention time

Computer Vision Research Engineer

February 2022 - May 2022

LENS Corporation, India

Corrected and evaluated programs for fingerprint matching algorithms in C++ using ROC Curve and similarity score metrics
Programmed a C++ pipeline with OpenCV and Torchscript for deploying deep learning models for fingerprint feature extraction on GPU and CPU systems

Perception Research Engineer

October 2020 - February 2022

Autonomous Vehicle Project | IIIT Delhi, India

Designed and deployed a robust system for calibrating 2 front-facing cameras and a LiDAR on a self-driving vehicle, enhancing sensor fusion accuracy
Benchmarked and optimized lane detection and tracking algorithms using Torchscript and C++, integrated with ROS for autonomous vehicle deployment, achieving 12+ FPS real-time processing

Research Intern

February 2020 - October 2020

Microsoft Research India

Led the deployment of the Automated Driver License Testing (ALT) project at the Institute of Driving Training and Research (IDTR) in Bihar
Implemented algorithms to refactor and scale the project for various Regional Transport Offices (RTO) across India, leading to deployment in 10+ cities with 99% accuracy