Challenge
Recent work revealed that state-of-the-art text recognition methods perform well on images with words within vocabulary, but generalize poorly to outside vocabulary text images. In real-world scenarios out-of-vocabulary (OOV) words are common and of great importance, for example, emails, dates, random strings. Unfortunately, existing benchmarks do not contain many of these words, and therefore, current methods are not evaluated on OOL and overweighting their own (implicit and explicit) language model.
This workshop proposes The OOV Challenge (OOV-ST), presenting an evaluation set of OOV text-in-the-wild images. This challenge attracts vision-language combined methods that are robust to OOV words. We expect this challenge to increase interest in techniques that balance the trade-off between vision and language.
To download the dataset and join the challenge, please visit the RRC portal.
Challenge Organisers
Sergi Garcia Bordils
Computer Vision Center, Autonomous University of Barcelona
Andres Mafla Delgado
Computer Vision Center, Autonomous University of Barcelona
Ali Furkan Biten
Computer Vision Center, Autonomous University of Barcelona
Lluis Gomez
Computer Vision Center, Autonomous University of Barcelona
Ruben Perez Tito
Computer Vision Center, Autonomous University of Barcelona
Ron Litman
AWS AI Labs
Aviad Aberdam
AWS AI Labs
Shai Mazor
AWS AI Labs
Shangbang Long
Google Research
Siyang Qin
Google Research
Amanpreet Singh
Huggingface
Computer Vision Center, Autonomous University of Barcelona