id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_vft4ua43ojg5dj5wcu6e4d43ka	Iam Palatnik de Sousa	Convolutional ensembles for Arabic Handwritten Character and Digit Recognition	2018	13	.pdf	application/pdf	5545	503	55	dataset, also yielding state of the art validation and testing classification accuracies: How to cite this article Palatnik de Sousa (2018), Convolutional ensembles for Arabic Handwritten Character and Digit Recognition. on a dataset of 6,600 images of characters, obtaining a validation accuracy of 97.32%. namely, the selection and preparation of the datasets, the network architecture, the training The datasets chosen for training and parameter tuning were the MADbase and AHCD networks, each trained with two different strategies (with data augmentation and without). It was also observed that the final averaged test accuracy of 6-fold validation for MADbase validation and test accuracies for MADbase. The average test and validation accuracy values of ENS4 are very promising and improve The system was trained and tested on the two largest available datasets of Arabic digits and and one that ensures the test and validation sets have the same size for the MADbase dataset.	./cache/work_vft4ua43ojg5dj5wcu6e4d43ka.pdf	./txt/work_vft4ua43ojg5dj5wcu6e4d43ka.txt