// welcome
I work on
I am Hussain, an undergraduate researcher working at the intersection of perception and intelligent systems. I build things that help machines see, understand, and act in the physical world.
My primary focus is computer vision ranging from classical image processing to deep learning-based detection and segmentation. I am fascinated by how visual understanding can be paired with robotics to create autonomous systems that navigate, manipulate, and make decisions in real environments. Lately, I have been diving deep into applied ML, particularly vision-language models (VLMs) that bridge the gap between seeing and reasoning. I have also been putting these skills to work in precision agriculture using drones, LiDAR, and remote sensing to phenotype crops and automate field analysis.
YOLO-World-powered robot that classifies rooms on sight, skips irrelevant ones, and remembers layouts across missions, distributed across Pi 5 and laptop.
Zero-install web app for ROS1 (.bag) and ROS2 (.mcap, .db3) files with 3D point clouds, time-series plots, and TF-aware rendering, fully client-side via Web Workers and WASM.
Incremental SfM pipeline reconstructing 200K+ 3D points from 30 images with bundle adjustment, served as an interactive Three.js virtual tour.
Robomaster robot with computer vision and PID control to detect markers, pick up objects, and place them at target locations.
YOLOv4 deep learning to detect EV signs, charging stations, and accessible parking signs achieving 74% mAP via transfer learning.
Turtlebot3 in ROS1/Gazebo detecting obstacles, mapping positions, and generating 2D environment visualizations.
Infrastructure-free UAV phenotyping using only RGB imagery and a consumer drone. Dual-altitude RANSAC-corrected orthomosaics achieve 94.95% plant counting and 91.59% panicle detection accuracy, with strong DEM-ground truth correlation (R = 0.876) — all without GCPs or RTK-GPS.