Introduction

This repository contains the necessary materials to reproduce the results I obtained in my report.

Package requirements

The codes should be run in Python 3. The following packages will be required to run the codes (latest versions should work):

• numpy
• matplotlib
• tqdm
• tensorflow
• scikit-image

All packages can be installed with $pip intall name_of_the_package.

User manual

1. Download the folders called shapes_6dof, slider_depth and office_zigzag (http://rpg.ifi.uzh.ch/datasets/davis/shapes_6dof.zip ; http://rpg.ifi.uzh.ch/datasets/davis/slider_depth.zip ; http://rpg.ifi.uzh.ch/datasets/davis/office_zigzag.zip) and put it in data.
2. Run $python txt_to_npy.py to convert .txt files to .npy files.
3. Run $python gradient_descent.py to perform gradient descent (on slider_depth by default).
4. Run $python ADMM.py to perform ADMM (on slider_depth by default).

Content

• data contains the file downloaded from http://rpg.ifi.uzh.ch/davis_data.html (under Text (zip)). The files must be downloaded before running the codes. Once you have downloaded a folder, put it in data.
• events will contain the stream of events under .npy format.
• dataio.py manages the data loading.
• txt_to_npy translates .txt files to .npy files.
• gradient descent.py contains experiments with gradient descent.
• ADMM.py contains experiments with ADMM.

Modifying the experiments

• To use another stream of events than slider_depth, just uncomment the indicated lines in gradient_descent.py or ADMM.py

# Uncomment to use shapes_6dof
# ************
Uncomment this
#************

• The parameters of the reconstruction can be modified under the line

### RUN EXPERIMENT ###

Have fun!