3D Reconstruction - Multiple Viewpoints by Columbia University
This course focuses on recovering a scene's 3D structure from photos shot from various angles. They begin by creating a detailed geometric model of a camera, and then devise a method for determining (calibrating) the camera model's internal and external parameters. Then they show how two calibrated cameras with known relative locations and orientations can be utilized to recover the scene's 3D structure. Simple binocular stereo is what they call this. The challenge of uncalibrated stereo, in which the relative positions and orientations of the two cameras are uncertain, is next. They can both calculate the relative positions and orientations of the cameras and then utilize this information to estimate the 3D structure of the scene using only the two images taken by the cameras.
The problem of dynamic scenes is then addressed. They describe how to compute the motion of each point in an image given two photographs of a scene with moving objects. Optical flow refers to the apparent motion of points in a picture. They can track scene points across a video stream using optical flow estimation. Next, They will look at a video of a scene captured with a moving camera whose motion is uncertain. They offer structure from motion, which uses tracked characteristics in a video as input to determine not only the scene's 3D structure, but also how the camera moves in relation to the scene. Object modeling, 3D site modeling, robotics, autonomous navigation, virtual reality, and augmented reality all leverage the methods we develop in this course.
This course offers:
- Flexible deadlines: Reset deadlines in accordance to your schedule.
- Certificate : Earn a Certificate upon completion
- 100% online
- Beginner Level
- Approx. 8 hours to complete
- Subtitles: English
- Course 4 of 5 in the First Principles of Computer Vision Specialization
Coursera Rating: 4.6/5
Enroll here: https://www.coursera.org/learn/3d-reconstruction-multiple-viewpoints
