Structure from Motion Quiz

Answers and explanations appear as you take the quiz.

1. What two things does Structure from Motion estimate simultaneously?
   • 3D point locations and the camera positions that took the photos
   • Image colors and brightness levels
   • Only the camera's battery life
   • The names of objects in the scene
2. What is the role of feature matching in SfM?
   • Adding color to the final model
   • Identifying the same physical point across multiple overlapping images
   • Compressing the images for storage
   • Deciding the camera's shutter speed
3. What does bundle adjustment optimize?
   • The file size of the output
   • The order in which photos were taken
   • Reprojection error across all cameras and 3D points at once
   • The contrast of each image
4. What kind of 3D output does SfM typically produce directly?
   • A fully textured, watertight solid mesh
   • A 2D thumbnail
   • An audio waveform
   • A sparse point cloud plus calibrated camera poses
5. Which geometric concept relates corresponding points between two camera views?
   • Epipolar geometry
   • Fourier transform
   • Color quantization
   • Backpropagation
6. Why is SfM important for modern methods like NeRF and Gaussian Splatting?
   • It colorizes the NeRF output
   • Those methods need accurate camera poses, which SfM provides
   • It replaces the need for any images
   • It compresses the neural network weights
7. Given a matched image point in one camera view, what does epipolar geometry constrain in a second view?
   • It must have the same pixel coordinates
   • It must lie at the image center
   • Its corresponding point must lie on a particular epipolar line
   • It uniquely determines the point’s color
8. In bundle adjustment, what is a reprojection error?
   • The number of images in the dataset
   • The angle between two RGB channels
   • The GPU memory needed for the point cloud
   • The pixel distance between an observed feature and the projection of its estimated 3D point