Things I've built, broken, and occasionally finished. Mostly at the intersection of robotics and computer vision.
Clean C++ Implementation of Zhang's Calibration method.
Infrastructure-based 3D perception stack for roadside autonomy, BEV tracking, multi-camera projection, and traffic risk analytics
A camera-only perception pipeline that turns KITTI driving video into metric depth, road-aware BEV occupancy, tracked object footprints, and object-aware risk heatmaps.
Interactive 3D LiDAR point cloud and infrastructure camera image visualizer for TUMTraf-V2X.
Weakly supervised RGB camera reliability scoring for driving scenes using OctoSense captions, ResNet18, visual-statistics baselines, depth-error validation, video diagnostics, and Grad-CAM.
ReLU vs Leaky ReLU vs GELU in a PointNet baseline — accuracy, stability, and latency under a fair, single-switch setup
Tiny PyTorch codebase for playing with transformer attention on simple BEV (bird’s-eye-view) grids and seeing how the softmax temperature changes attention patterns and downstream predictions.
structure from motion - clean and readable code unlike all the other implementations
Hide in Image
Use yolov8 to track and reid the person in the frame
