Handwriting recognition is a challenging task and with the advancements in the development of the deep learning such task can be performed even for very limited documents. This paper aims to perform writer identification and retrieval from historical documents using an ensemble of convolutional neural network models that were built using the Inception-ResNet-v2 pre-trained architecture. The dataset comprises 170 images grouped in 34 classes. The results prove that the ensemble model outperforms single pre-trained models, obtaining an accuracy of 96%.
Writer Identification from Historical Documents Using Ensemble Deep Learning Transfer Models
Amelio A.Secondo
;
2022-01-01
Abstract
Handwriting recognition is a challenging task and with the advancements in the development of the deep learning such task can be performed even for very limited documents. This paper aims to perform writer identification and retrieval from historical documents using an ensemble of convolutional neural network models that were built using the Inception-ResNet-v2 pre-trained architecture. The dataset comprises 170 images grouped in 34 classes. The results prove that the ensemble model outperforms single pre-trained models, obtaining an accuracy of 96%.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
