Skip to content

Home

Jump to bottom
H Gazula edited this page Apr 29, 2023 · 7 revisions

Welcome to the photo-reconstruction wiki!

Project Github link: here

Path on calico: /space/calico/1/users/Harsha/photo-reconstruction

Setting up the project:

Please refer to the setup script.

Python Environment:

  • Option 1. Build your own using the requirements file.
  • Option 2. Source my environment- source /space/calico_001/users/Harsha/venvs/recon-venv/bin/activate
  • Option 3: Source freesurfer source /usr/local/freesurfer/nmr-dev-env-bash

NOTE 01: If you source freesurfer, use fspython instead of python wherever appropriate.

NOTE 02 (important): Before running this project, please add the project directory to the PYTHONPATH environment variable as follows:

export PYTHONPATH=/space/calico/1/users/Harsha/photo-reconstruction

Directory Structure: (followed by a description of contents)

image

data

  • 4harshaHCP - HCP data to quantify reconstruction error on synthetic data.
  • prob_atlases - Atlas templates for the population. Copy of /cluster/vive/prob_atlases
  • uw_photo - Subset of /cluster/vive/UW_photo_recon

ext - External modules used in the project

  • lab2im - utils.py is an exact copy of lab2im/utils.py from SynthSeg. Some functions that are dependent on tensorflow/keras have been commented out.

init.sh - Shell script to setup the project in your home directory

trash - I often maintain a trash directory in every project to avoid accidents.

photo_reconstruction

  • LBFGS.py - Optimizer for the reconstruction model.
  • versatile_reg_network.py - Pytorch model with reconstruction objective

requirements.txt - Requirements file for building the environment

results

  • hcp-results-20220912 - The most up-to-date results from the HCP simulation.
  • uw_gt_map.csv - Index of the ground truth slice for UW data (1-indexing for MATLAB). For more information, please refer to this page
  • output_gt_slice_idx_hard.txt - Index of the ground truth slice for UW data (0-indexing for Python)

scripts

  • 3d_photo_reconstruction.py - A local copy* of mri_3d_photo_recon but with extra flags to simulate various skips. *Currently, this script is lagging mri_3d_photo_recon (but not too far behind).
  • hcp_replicate_photos.py - Script to simulate the HCP experiment. Please take a look at experiment 1 in the paper for more information
  • hcp_registration_error.py - Script to calculate and plot (boxplot) reconstruction error on the data simulated in the above script. Before running this script, you have to run make hcp_recon. Please refer to the Makefile for more information.
  • fig-recon-error.ipynb - This notebook is a modified copy of the plotting function from hcp_registration_error.py to plot the final figure used in the manuscript.
  • *.m - For information on the Matlab scripts, please talk to Eugenio. He can explain them better.

submit.sh - SLURM submission script

Clone this wiki locally