Transform raw 3D spatial data into accurate, production-ready geometric models and perception software. I specialize in developing custom 3D Computer Vision workflows, LiDAR point cloud algorithms, and 3D deep learning pipelines using Open3D, PyTorch3D, and OpenCV.
What I Offer
- Point Cloud Processing & Filtering: Downsampling (voxel grid), noise removal, outlier filtering, and surface normal estimation.
- 3D Registration & Alignment: Global alignment and local registration (ICP / Iterative Closest Point) for scan merging and SLAM.
- Point Cloud Segmentation & 3D Bounding Boxes: Floor/ground plane removal, RANSAC geometric fitting (spheres, cylinders, planes), and 3D object detection.
- Mesh Generation & Reconstruction: Converting raw point clouds into clean 3D CAD meshes (Poisson surface reconstruction, Ball Pivoting).
- 3D Deep Learning: PointNet++, VoxNet, or MinkowskiEngine models for semantic segmentation and 3D classification