Loading...

Recognition of Persian Handwritten Digits Using Neural Network

Ghasemi Saghand, Esmat | 2021

244 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54661 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Rafiee, Majid; Mousavi, Mohsen
  7. Abstract:
  8. In this project, a convolutional neural network is trained to classify Persian handwritten digits and different neural network models are compared for this problem. To train and evaluate this network, the "HODA" data set has been used, which includes 80,000 images of Persian handwritten digits, including 60,000 images of training data and 20,000 images of test data. In the modeling section, we have used a comparison of recent top architectures of convolutional neural networks to identify Persian and English handwritten numbers. In this experiment, the Tensorflow library in the Python programming language was used. Parameter values and network hyper parameters are obtained from the values specified in recent top articles. Also, according to the results of different models, we designed a model that has a very good percentage of accuracy and error on our collection, which is equal to the accuracy of the training data set 99.98% and the accuracy of the validation data set 99.77% and the accuracy of the test data set 99.45% on Persian handwritten numbers. The accuracy of the superior model of this study is higher than the accuracy mentioned in the articles that have used the HODA Persian handwriting numbers only by using the convolutional neural network method, and the model works excellently
  9. Keywords:
  10. Artificial Intelligence ; Deep Learning ; Convolutional Neural Network ; HODA Dataset ; Persian Handwriting ; Persian Handwritten Digits Recognition

 Digital Object List

 Bookmark

No TOC