Nakba Image Classification

This Shared Task is part of the Nakba-NLP 2026 Workshop, itself part of LREC 2026.

Motivation

Since October 7, 2023, the Tech for Palestine incubator project SaltPillar has been archiving millions of images from social media.

Some of these images pertain directly to the violence of the Nakba: airstrikes, shooting, injury, death, blood, terror, grief, destruction. Other images are only indirectly relevant: children playing, adults drinking coffee, journalists giving speeches abroad, etc. As we move to make these data more accessible, categorizing the data is of high priority.

Data Description

• Data Source: Instagram
• Dataset Size: 2,001 images
• 706 images showing destroyed/damaged infrastructure
• 1,295 images not showing destroyed/damaged infrastructure
• Scope: Images published by Palestinian content creators or journalists in Gaza from October 7th, 2023 until December 15th, 2025
• Data Fields:
  • Image (.png)
  • Label (a short string categorizing the image)

Task Definition

In this Shared Task, participants will be asked to train a binary image classifier on 2,001 images, to help distinguish images of infrastructure that is destroyed/damaged from infrastructure that is intact. The classifier will ultimately be tested on a hold-out set of 500 images.

Example of an image showing destroyed/damaged infrastructure:

Example of an image showing infrastructure that is still intact:

Tentative Schedule

• January 1st, 2026: Call for Participation
• January 10th: Release of Training Data (2,001 images and labels).
• February 10th: Evaluation Period Begins (Test set: 500 images)
• Feb 17th: Evaluation Period Ends.
• Feb 21st: Release of Results.
• Mar 1st: Paper Submission Deadline.

Organizers

• Alexei Abrahams, Hamad bin Khalifa University
• Shadi Abudalfa, King Fahd University
• Mustafa Jarrar, Hamad bin Khalifa University
• George Mikros, Hamad bin Khalifa University

Sign up!

To sign up for this Shared Task, please fill out the Google form.