Abstract
The photo document image's quality determines whether it has the potential to be used for information extraction. Document Image Quality Assessment (DIQA) is a difficult task since it is complicated to train a system to have a complete human-like vision and it might be tiresome and time-consuming to manually evaluate the quality of many document images. The educational system became online during the pandemic age, which led to the digital submission of exams and homework. Quality evaluation has gained more prominence recently as digitization has started to take precedence in some fields. In order to evaluate the document image quality of handwritten document images at the page level while taking into account the overall visual aesthetic, this research suggests a transfer-learning and classification-based model. ResNet50 exceeded all of the pre-trained models we tested and examined, with an accuracy of 91.00 % on validation sets and 83 % on test sets.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call