Skip to content
Experiments to measure the behavior of deep neural networks performing a visual search task.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


Experiments to measure the behavior of deep neural networks performing a visual search task.

For some background and a summary of the results, please see this Jupyter notebook.


Experiments were run using Anaconda on Ubuntu 16.04. The following commands were used to create the environment:

tu@computi:~$ conda create -n searchnets python=3.6 numpy matplotlib imageio joblib tensorflow-gpu 
tu@computi:~$ source activate searchnets
tu@computi:~$ git clone
tu@computi:~$ cd ./visual-search-nets
tu@computi:~$ pip install .


Installing this package (by running pip install . in the source directory) makes it possible to run the experiments from the command line with the searchnets command, like so:

tu@computi:~$ searchnets train config.ini

The command-line interface accepts arguments with the syntax searchnets command config.ini,
where command is some command to run, and config.ini is the name of a configuration file with options that specify how the command will be executed.
For details on the commands, see this page in the docs. For details on the config.ini files, please see this other page.


Data is deposited here:

Replicating experiments

The Makefile replicates the experiments.

tu@computi:~$ make all


  • Research funded by the Lifelong Learning Machines program, DARPA/Microsystems Technology Office, DARPA cooperative agreement HR0011-18-2-0019
  • David Nicholson was partially supported by the 2017 William K. and Katherine W. Estes Fund to F. Pestilli, R. Goldstone and L. Smith, Indiana University Bloomington.
You can’t perform that action at this time.