Visual SLAM for Robotics Build a VSLAM system in Python
Published 8/2026
Created by Ferbin Richard
MP4 | Video: h264, 1920×1080 | Audio: AAC, 44.1 KHz, 2 Ch
Level: All Levels | Genre: eLearning | Language: English + subtitle | Duration: 61 Lectures ( 41h 38m ) | Size: 10.8 GB
Build a complete Visual SLAM system in Python & ROS 2 — feature tracking, bundle adjustment, loop closure and 3D map
What you’ll learn
Build a complete monocular, stereo and RGB-D visual SLAM system in Python from scratch — feature tracking, pose estimation, mapping and loop closure.
Master the geometry behind VSLAM: camera calibration, epipolar constraints, the essential and fundamental matrices, triangulation and PnP pose recovery.
Implement a full back end — keyframe selection, local and global bundle adjustment, pose-graph optimization and drift correction with g2o/GTSAM-style solvers.
Run and tune ORB-SLAM3, RTAB-Map and visual-inertial odometry on real robots in ROS 2, and fuse IMU data for scale and robustness.
Requirements
Basic Python. If you can write a function and a loop, you are ready — every algorithm is built step by step from first principles.
High-school linear algebra is enough. Matrices, vectors and transforms are re-taught from scratch as the course needs them.
A laptop with Ubuntu (or WSL2 / Docker on Windows). No robot, no depth camera and no GPU required — everything runs in simulation and on free datasets.
Description
This course contains the use of artificial intelligence.
Every autonomous robot has to answer two questions at the same time
Where am I, and what does the world around me look like?
Visual SLAM, or Simultaneous Localization and Mapping using cameras, is how robots solve both. It is used in drones, warehouse robots, AR headsets, autonomous vehicles, and many other systems that need to understand motion and build maps without relying entirely on GPS.
This course takes Visual SLAM apart and builds it back up from first principles.
Build Visual SLAM from scratch
You will create a working SLAM pipeline in Python, one component at a time.
You will start with
Camera calibration
ORB feature detection and matching
Essential matrix estimation
RANSAC
Relative camera motion
3D point triangulation
PnP pose estimation
Visual odometry
You will watch your own camera trajectory appear on screen from the system you built.
Turn visual odometry into SLAM
Next, you will add the components that make it a complete mapping system
Keyframes and map points
Local and global bundle adjustment
Pose graph optimization
Bag-of-words place recognition
Loop closure
Tracking failure detection
Relocalization
You will also understand one of monocular SLAM’s biggest limitations:scale ambiguity, and learn how stereo and RGB-D cameras solve it.
Add IMUs and ROS 2
You will then move from pure visual SLAM into robotics applications
Stream camera data through ROS 2
Fuse camera and IMU measurements
Understand visual-inertial odometry
Compare your implementation with ORB-SLAM3 and RTAB-Map
Connect SLAM outputs to a robot navigation stack
Benchmark like a robotics researcher
You will evaluate SLAM performance using public datasets including
KITTI
EuRoC
TUM RGB-D
Instead of deciding whether a trajectory “looks good,” you will measure localization and trajectory error using the same ideas commonly used in robotics research.
No expensive hardware required
Everything can be completed using
A normal laptop
Simulation
Free public datasets
Python and ROS 2
No robot, depth camera, or GPU is required.
By the end of the course, you will not just know what Visual SLAM is. You will havebuilt one yourself.
You will understand front ends, back ends, keyframes, bundle adjustment, loop closure, visual-inertial odometry, and the mathematical ideas underneath them.
Most importantly, you will be able to look at a drifting trajectory or broken map and reason aboutwhy it failed and how to debug it.
Who this course is for
Robotics engineers and students who can use SLAM as a black box but want to understand and modify what happens inside it.
Computer vision developers moving into robotics who want to apply feature matching and multi-view geometry to a real navigation stack.
ROS 2 developers who need reliable localization and mapping from cameras instead of expensive 3D LiDAR.
HVIALLAMFORROBOTICBUILDAVLAMYTE
you must be registered member to see linkes Register Now