Skip to content

Code for our entry into the 2022 CompSAN Data Competetion

Notifications You must be signed in to change notification settings

JunaidMerchant/CompSAN2022_DCteam

Repository files navigation

CompSAN 2022 Data Competition - Washington DC Team

This was a collaborative effort by Junaid S Merchant, Shawn A Rhoads, Oliver Xie, and Sarah Dziura (Georgetown University and University of Maryland).

  • Please note that because this was a collaborative effort across different institutions and computing resources, the paths and formats of the different pieces of code are not unified (reflecting different practices of the people involved.)

  • Though this repository does not include code for all the different analyses that we conducted (not included in the abstract or poster). However, along the way, we relied heavily on python libraries (nilearn, sklearn, nibabel, nltools, scipy, networkx, and others), MATLAB toolboxes, (cvVectorStats - https://github.com/fwillett/cvVectorStats, CoSMoMVPA - https://github.com/CoSMoMVPA, AlexNet, and others), as well as numerous AFNI functions.

Brief descriptions of the scripts

/actor_images - directory with images to derive alexnet features

alexnet_feature_extraction.m - code used for extracting alexnet features

aws_rekognition_get_celebrities.md - describes basic AWS Celebrity Detect Rekognition functions

aws_rekognition_page_results.sh - code for paging through rekognition results and printing to json files

aws_rekognition_parse_celebrity_output.py - parsing the resulting json files from rekognition

/celebrity_detect_rekognition_data - data from rekognition that includes time stamps for each actor

compSAN_DC_Group_Abstract.pdf - abstract submitted to CompSAN data competition

compSAN_DC_Group_Poster.pdf - poster submitted to CompSAN data competition

cosmo_pdist_jsm.m - modified cosmomvpa function

cosmo_target_dsm_corr_measure_jsm.m - modified cosmomvpa function

/event_timing - directory with face event timing and regressor files for first-level modeling

get_wikinet_rdms.py - construct RDM from edgelist generated from Wikipedia pages and plot figures

grouplevel_searchlight_alexnet.sh - group-level t-test on subject searchlights for alexnet rsa

grouplevel_searchlight_social.sh - group-level t-test on subject searchlights for social relationship rsa

lss_conca_timing.m - concatenate timing for afni lss

lss_conca_timing_v2.m - concatenate timing for afni lss

lss_covert_1D.m - convert timing to 1d files for afni lss

lss_covert_1D_v2.m - convert timing to 1d files for afni lss

merge_betamaps.sh - merge beta-maps for ease of use with cosmomvpa

/model_rdm_data - csv files of rdm data used in rsa

remove_zero_cosmo_ds.m - helper function for use with cosmomvpa that removes zeros before calculating similarity

run_3dLSS.sh - script for running afni 3dLSS

run_alexnet_rsa.sh - wrapper for running subject-level alexnet searchlight on high performance computer

run_social_rsa.sh - wrapper for running subject-level social relationships searchlight on high performance computer

searchlight_rsa_alexnet.m - main function for running subject-level alexnet searchlight using cosmomvpa

searchlight_rsa_social.m - - main function for running subject-level social relationship searchlight using cosmomvpa

setup_3dLSS.sh - 3ddeconvolve function used to set up xmats for afni 3dLSS

About

Code for our entry into the 2022 CompSAN Data Competetion

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages