FUNSD
A widely used benchmark dataset in document information extraction. It contains scanned form images with annotations covering text regions, semantic field roles, and inter-field relationships. The dataset is characterized by a relatively high level of OCR noise.
Timeline (last 90 days)
Different datasets such as FUNSD, CORD, and DocVQA do not have unified definitions of annotation granularity and boundaries.
The study uses three document extraction benchmark datasets: FUNSD, CORD, and SROIE.
FUNSD (Form Understanding in Noisy Scanned Documents) is a dataset specifically designed for form understanding.
All Facts (3)
Different datasets such as FUNSD, CORD, and DocVQA do not have unified definitions of annotation granularity and boundaries.
50%UnverifiedThe study uses three document extraction benchmark datasets: FUNSD, CORD, and SROIE.
50%UnverifiedFUNSD (Form Understanding in Noisy Scanned Documents) is a dataset specifically designed for form understanding.
50%