ASSIGNMENT 2: Bundle Adjustment
In this assignment, we will learn to simultaneously compute the extrinsic parameters of a set (bundle) of cameras, given multiple pairs of relative transformations between them. It’s an instance of the bundle adjustment problem, which is very important for stereoscopy and 3D reconstruction. We’ll use PyTorch3D API to set up an optimization process and minimize the discrepancies between pairs of relative cameras.
The goals of this practice are the following:
- Get more familiarization with camera representations
- Discover new APIs in PyTorch3D to work with rigid transforms
- Use the power of neural networks and gradient descent to solve a classic optimiation
![](img/random_camera_bundle.png)
Instructions:
- Follow the instructions in the notebook for completing the assignment.
Assignment 2 Notebook
![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)
Submission:
The assignment is due on March 30th, 2024 at 11:59pm (EDT).
References:
- Jonas Gomes, Luiz Velho and Mario Costa Sousa. Computer Graphics: Theory and Practice, chapters 2, 3 and 4.
- Richard Hartley and Andrew Zisserman. Multiple View Geometry in Computer Vision, chapter 18.
- PyTorch3D Tutorials - Bundle Adjustment