|
In this work, we present a new dataset called Tifinagh-MNIST: handwritten letters of the Tifinagh alphabetic script, which is used to write the Tamazight languages. The presented dataset contains 82,500 gray scale images of size 28 x 28 pixels belonging to 33 classes (letters), with 2500 images per class. In particular, the training set is composed of 66,000 images while the test set contains 16,500 images. Tifinagh-MNIST is intended for the development of AI tools to process handwritten Tifinagh characters. We also propose use cases of this corpus through neural network models for classification and data generation. The corpus will be made available to the public in order to promote the development of artificial intelligence solutions for the processing of handwritten Tamazight texts. |
