Primary competition visual

R.O.A.D. Barbados Historic Handwriting Challenge

Helping Barbados
$25 000 USD
~2 months left
Optical Character Recognition
Natural Language Processing
1218 joined
272 active
Starti
03 Jul 26
Closei
04 Oct 26
Reveali
04 Oct 26
Organizer ruling needed: Test-image pseudo-label self-training
4 Aug 2026, 12:42 · 0

Hi Zindi team, I am reopening this clarification because the earlier question, "Rules clarification: self-training / pseudo-labeling on the test images," remains without an organizer response: https://zindi.world/competitions/road-barbados-historic-handwriting-challenge/discussions/34110 Could an organizer please provide a clear yes/no ruling: May participants use the challenge Test images to generate pseudo-labels entirely in code and then continue training or fine-tuning a model on those Test images, provided that no external data, external labels, or manual labeling is used? The rules permit openly available pretrained models and challenge-provided datasets, but do not clearly authorize transductive Test-image training. Until Zindi confirms this in writing, we will treat Test images as inference-only. A written ruling would help participants avoid an unintended rules violation. Thank you.

Discussion 0 answers