This project was done for Robomath (16811) @ CMU in Fall 2022. All the bookkeeping, initialization, and optimization (including analytical Jacobians!) were done by hand in python. Usage as follows:

• run.py runs SfM. Run python run.py --help for input/output options. For example, after downloading the moose data (see line below), it can be ran quickly with the first 15 images at 50% scale and live plotting with python run.py -p -s 25 -n 15. Note live plotting is not interactive until SfM has finished, and output checkpoints will be saved to results/ by default. Also, currently the initial estimates for the intrinsics are set inside the file.
• download_data.py downloads and extracts the data for the moose ornament example into the data/ folder. Make sure you install dependencies for this first (see below).
• plot.py shows the 3D point cloud. Run python plot.py results/point_cloud.npz to see the resulting moose ornament point cloud.
• slide_images.py is full of some helpers to create images for the slides/writeups.
• point_cloud_to_mesh.py was a first attempt at some surface reconstruction from the dense point cloud. Requires more tuning.

The majority of the SfM code can be found in the sfm directory.

More technical info can be found in the writeup and slides.

I recommend installation into a conda environment. A simple command to make and activate a new environment is

conda create -n sfm python=3.10
conda activate sfm

Then to install all required dependencies, run

pip install -r requirements.txt