Skip to content

Latest commit

 

History

History
25 lines (14 loc) · 1.36 KB

README.md

File metadata and controls

25 lines (14 loc) · 1.36 KB

StyleGAN2 Hypotheses Explorer

This repository implements the StyleGAN2 Hypotheses Explorer, a framework to interactively explore an image classifier's decision boundary. This project is associated with our paper Explainability Requires Interactivity. For training of networks and other stuff from the paper, consider the sister repo.

Setup

  1. Setup the server.
  2. Setup the client.

Run

  1. Start the server.
  2. Start the client.

Export

The StyleGAN2 Hypotheses Explorer can be exported into a format which only requires a file server (at the moment it can only be served from the root directory of a domain. See here on how to configure serving from a subdirectory manually).

  1. Configure the export process by creating an export_settings.json file as described here.
  2. Activate the stylegan2_hypotheses_explorer environment via conda activate stylegan2_hypotheses_explorer.
  3. Run python export.py path/to/export_settings.json.

After running the export script the exported version (including client and server) can be found in the directory specified in the export_settings.json.