Dilip G.

Computer Vision Engineer | 3D Reconstruction | RGB-D & Sensor Fusion M.Sc. in Geodesy and Geoinformation Science (TU Berlin), majoring in Photogrammetric Computer Vision and Remote Sensing. I help teams turn images, video, depth, and sensor data into reliable computer vision and 3D perception systems. My experience covers 3D reconstruction, RGB-D processing, geometric registration, deep learning, medical imaging, mobile vision, AR, and privacy-aware visual analysis. I've worked as a research trainee at Institut Pascal (EnCoV) and as a computer vision intern at SurgAR, both in Clermont-Ferrand, France, building complete workflows — from data preparation and model development to quantitative evaluation, APIs, testing, and deployment. Core Expertise: 3D Vision & Geometry RGB-D and TSDF reconstruction, point clouds, meshes, camera calibration, multi-view geometry, Structure from Motion, Kabsch alignment, RANSAC, ICP, pose estimation, and monocular depth estimation Deep Learning & Computer Vision PyTorch, CNNs, Vision Transformers, YOLO, U-Net++, object detection, semantic segmentation, classification, landmark estimation, transfer learning, and model evaluation Software Engineering Python, C++, Swift, OpenCV, Open3D, Docker, FastAPI, TensorRT, Git, CI/CD, multiprocessing, automated testing, and API development Selected Results: Leading development of a 3D anatomical landmark estimation pipeline for the human ear (pinna), using PointNet and DGCNN heatmap models over full head-and-torso meshes — achieving 1.039 mm mean landmark error with subject-level cross-validation to prevent identity leakage. Built an iOS ARKit system for synchronized RGB, depth, confidence, camera-intrinsic, and pose capture, with a Python/Open3D reconstruction backend. Designed monocular 3D facial reconstruction and personalized geometric-analysis workflows from mobile imagery. Created resumable multiprocessing pipelines for video extraction, validation, metadata generation, and structured AI dataset preparation, supporting segmentation, viewpoint classification, landmark estimation, synthetic data generation, and visual anonymization. I can support projects involving 3D reconstruction, medical or general computer vision, geometric processing, model development, dataset engineering, technical evaluation, and vision API implementation. My priorities are clean engineering, measurable performance, reproducibility, and solutions that are easy to test and extend.