This GitHub repository contains machine learning models for character recognition in multiple languages, including MNIST, Arabic, Japanese. We have implemented various models such as Convolutional Neural Networks (CNN), Echo State Networks (ESN), and a combination of CNN-ESN & CNN-EuSN for classification purposes.
-
CNN Model: This model utilizes Convolutional Neural Networks to perform character recognition.
-
ESN Model: The Echo State Network model is another approach for character recognition.
-
EUSN Model: The Euler State Network (ESN) model is a powerful approach for character recognition. Harness the potential of ESNs to unlock accurate and efficient character classification.
-
CNN-ESN Hybrid Model: We have also developed a combined model that leverages the strengths of both CNN and ESN. This hybrid model aims to achieve improved accuracy and performance in character classification tasks.
-
CNN-EUSN Model: The CNN-Euler State Network (ESN) model combines the strengths of Convolutional Neural Networks (CNN) and ESNs for robust and accurate character recognition. This hybrid model, implemented as a class, offers a seamless integration of both CNN and ESN techniques, enabling you to leverage the power of deep learning and dynamic temporal modeling. Experience the best of both worlds with the CNN-ESN model for advanced character classification tasks.
All Models have the naming convention of "Language_Model-Architecture.py". Each of the models utilize different data-pipelines and processing techniques. The Model architecture for ESN and EUSN of each language is the same. CNN architectures for each of the languages have changed due to varied datasets and complexities of languages.
Note - The individual architecture of the models will be available soon for further usage.
To use any of the Models or Model architecture, you can add change the path of the datasets that are being imported and run the model and save the same in .h5 format. If any issue arises during running kindly flag the same in the "Issues" coloumn of this repo. Note - These are not pre-trained models, you have to train for your own usage.
We welcome contributions from the community to enhance the existing models or add support for additional languages. If you have any suggestions, bug reports, or feature requests, please feel free to open an issue or submit a pull request. Together, we can improve the accuracy and versatility of character recognition models.