The Signal Classification and Reconstruction Network (SCRNet) is a deep neural network architecture designed to handle compressed and noisy character images signals. Tailored for tasks like character recognition and image signal restoration, SCRNet integrates classification and reconstruction pathways, enhancing performance and robustness through their synergistic interaction.
Jupyter notebooks should be run in the following order::
- create_comp_noisy_emnist_letters_training_data.ipynb: Create training and test data
- scrnet_train_models_emnist_tf.ipynb: Train architectures using training data
- scrnet_model_analysis.ipynb: Analysis of models
