Discord

Checklist ROAI 2026 Selection Camp – GPU Practical Round · Task 3

Fake it until you make it

Detect Stable Diffusion v1.5 images among arbitrary real photographs when the training set contains only real images of dogs.

  • Vision
  • Binary image classification (AI-generated image detection)

The task

The training set consists exclusively of real photographs of dogs; no generated images are provided. The hidden test set contains random real images from the internet and images generated with Stable Diffusion v1.5, not necessarily of dogs.

The starter kit shows how to generate images with Stable Diffusion v1.5 and includes offline diffusers documentation and examples. The contestant predicts 0 (real) or 1 (AI-generated) for each test image.

Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.

At a glance

You get
Training images of real dogs; unlabelled test images; starter kit with a Stable Diffusion v1.5 generation example and offline diffusers documentation.
You submit
CSV with subtaskID (always 1), datapointID (test image id) and answer (0 or 1).
Scoring
Weighted F1: below 0.54 gives 5 points, 0.54–0.94 is scaled linearly from 5 to 100, 0.94 or above gives 100.
Rules
  • GPU round
  • On-site at the National College of Informatics "Tudor Vianu", Bucharest
  • Internet access and packages restricted to the allowed list in the round rules
  • Final score computed on hidden test data; only selected submissions count
Format
ROAI 2026 selection camp, GPU practical round, 27 May 2026.

Details

Year
2026, Bucharest, Romania (CNI "Tudor Vianu")
Round
Selection Camp – GPU Practical Round · Task 3
Language
English
License
Not stated by the source