This hands-on course introduces students to 3D computer vision using monocular and stereo cameras. Through a series of real-world projects and coding exercises, learners will build a strong foundation in camera geometry, feature-based matching, pose estimation, and 3D reconstruction targeted for research and industrial application in Autonomous vehicle, robotics, machine learning, 3d geometry and reconstruction.
You will begin by understanding camera calibration and how a single camera can be used for localization and height estimation. You’ll then move on to more advanced topics like real-time 3D pose estimation, augmented reality overlays, video stabilization, and visual odometry on real datasets like KITTI.
This course is project-driven and emphasizes classical, interpretable methods giving you the tools to develop your own computer vision pipeline without requiring deep learning.
What You Will Learn:
- Camera Calibration & Projection Geometry
- Estimate intrinsic and extrinsic parameters of monocular cameras
- Use projection grids for object height estimation
- Detect and track objects using feature matching
- Estimate 3D object pose and overlay augmented content in real-time
- Implement 2D video stabilization using feature tracking and homographies
- Perform planar image stitching using BRISK and homography transformation
- Use BRISK, ORB, and other descriptors for robust keypoint matching
- Understand outlier rejection using RANSAC
- Compute and visualize the fundamental matrix and epipolar lines
- Apply monocular visual odometry using optical flow and epipolar constraints
- Reconstruct 3D point clouds from stereo image pairs
- Understand triangulation using projection matrices
- All level python developers
