This directory contains files used to preform analysis on the data, and experiments on our models. Each of the scripts has a "-h" option for details.

Initial data analysis

generate_cluster_information_file.py: Generates a csv file with information about the clusters present in each image (such as number of clusters, size and intensity for each cluster). Precomputed file can be found in _____. To create the file from scratch, run python generate_cluster_information_file.py, followed by python generate_cluster_information_file.py --update_only

threshold_helper.py: Visualization tool we used to understand the effect of different intensity threshold values in determining which clusters in an image are detected as galaxies.

visualize_clusters.py: Visualization tool we used at the start of the project to visually understand the the almost invisible faint clusters in the images.

real_image_statistics.ipynb: A notebook where we perform data analysis on the extracted cluster properties. Depends on data/scored_info.csv

Experiments on generative models

analysis_patches_and_positions.ipynb: Notebook containing two experiments on the two-stage generative model: one to analyse how the position of the patches influences the score and one to analyse how the choice of patches influences the score.

Experiments on regressors

FFT visualize: Visualization tool for Fourier transform representation of images with different scores.