Human object learning benchmarks

This repository contains code for comparing models of human object learning against measurements of human behavior (n=371k trials).

Installation

We developed this repository using Python 3.9. Follow the steps below to use it:

1. Begin by cloning the hobj repository to your local machine.
2. Make and activate a new conda environment.
3. Install the hobj package itself. To do so, cd to the hobj directory (the top-level one), then run:

pip install -e .

4. Then, install the following dependencies:

conda install -c conda-forge xarray dask netCDF4 bottleneck

5. Install PyTorch. If you are using a computer without a GPU, you can run the command:

conda install pytorch torchvision torchaudio -c pytorch

Download all images (recommended)

The code in this repository works without this step. But to save time, it is recommended to download the images in a batch. They are located here.

Once it is downloaded, unzip it (it should turn into an images folder). Move that images folder to ~/hlb_cache/.

Usage

To see how to view the raw behavioral data and/or score an example learning model, check out the examples in examples/.