Spatial Frequency Maps in Human Visual Cortex:
A Replication and Extension

This repository includes codes to reproduce the analysis and figures to map spatial frequency preferences in human visual cortex using Natural Scenes Dataset - synthetic data.

for spatial frequency preferences in human visual cortex.

Citation

Ha, Broderick, Kay, & Winawer (2022). Spatial Frequency Maps in Human Visual Cortex: A Replication and Extension. .., .., . https://??

Table of Contents

• Dependencies
  • Conda environment
• Data
  • Processed data
  • NSD synthetic data
• Analysis pipeline
  • Reproducing the figures
  • Understanding the pipeline

Dependencies

All of the code in this repository is written in Python (3.7 and 3.8). To reproduce the python environment, we recommend using Conda to manage the dependencies.

Conda environment

1. Install miniconda for your system with the appropriate python version.
2. Install mamba: conda install mamba -n base -c conda-forge
3. Download this github repository: git clone [email protected]:JiyeongHa/Spatial-Frequency-Preference_NSDsyn.git
4. move to where the repository is downloded: cd path/to/sfp
5. run mamba env create -f environment.yml.
6. Type conda activate sfp to activate this environment and all required packages will be available.

Data

NSD synthetic data

The access to the NSD synthetic data will be granted after filling out the form on the NSD website (https://naturalscenesdataset.org/).

processed data

The data for this project is available on OSF (https://osf.io/umqkw/).

Analysis pipeline

Reproducing the figures

We used snakemake to manage the analysis pipeline. The pipeline is defined in the Snakefile in the root directory. To reproduce all the figures, you can use the following command: Add -N if you wish to run the pipeline in dry-run mode.

snakemake -j1 figure_all

Understanding the pipeline

We also provide a set of jupyter notebooks under the notebooks directory to understand the pipeline. The notebooks provide a step-by-step guide to the analysis pipeline. The number indicates the order of the analysis steps. Step 0: Stimulus check Step 1: Prepping the data Step 2: One-dimensional model: Fitting a log-Gaussian function to eccentricity-binned data Step 3: Two-dimensional model: Fitting the Broderick et al. model to the data