Skip to content

Web App for User Study in Django - (Image Generation Eval Metrics)

Notifications You must be signed in to change notification settings

cuiaiyu/user_study_interface

Repository files navigation

Human Study User Interface

This web app is developed for human study in Computer Vision research, especially for GAN related tasks. We implemented:

  1. Just Noticeable Difference (JND): show user an image for one second and ask real/fake
  2. A/B Test: Given reference images, ask user whether query 1 or query 2 has better quality.

How to start

This web app is implemented by Django. (A nice tutorial for anyone who is interested to read.)

Lanuch Website

python manage.py runserver 0.0.0.0:8000

Populate Question Pool

Manually

One way to do so is to manually add question as the superuser. First log as the superuser. Create superuser by:

python manage.py createsuperuser

Then go to 0.0.0.0:8000/admin to change question pools. You may also need to link answers to their associated questions.

Automatically

We are working writing a scrpt to do this. TBD.

Collect the statistics

Your users's report will be saved in the linked database. All you need to do is read the count of answers of every question, and evaluate according. We are working on developing a script which can do this automatically.


We borrowed some .css styles from https://github.com/akashgiricse/lets-quiz

About

Web App for User Study in Django - (Image Generation Eval Metrics)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published