Overview
Multimodal Foundation Models (MMFMs) have shown unprecedented performance in many computer vision tasks. However, on some very specific tasks like document understanding, their performance is still underwhelming. In order to evaluate and improve multimodal models for the task of document image comprehension, we harness a large amount of publicly available and privately gathered data and propose a challenge. Our challenge is running in two separate phases. For the first phase, we have released a training and a validation set consisting of publicly available data; and for the second phase, an alien test set will be released soon for the final leaderboard.
Dates
Phase 1 Submission Due: 15th May 2024
Phase 2 Submission Due: 30th May 2024
Solution Review: 1th-7th June 2024
Announcement of Winners: 8th June 2024
Sponsorship
Our challenge winners prize is awarded in collaboration with:
