Description:
This release introduces my initial Convolutional Neural Network (CNN) model tailored for Hindi OCR. With simple-yet-accurate architecture aiming to be a mobile model, the model is designed to capture spatial hierarchies and is highly efficient for image data.
Key Features:
- Tailored for Hindi character recognition.
- Utilizes multiple convolutional and pooling layers.
- Improved accuracy on validation dataset.
- Changes from Previous Version:
Initial release. Model Performance Metrics: Training accuracy: 94% | Accuracy on unseen samples: 97%
Dependencies:
- _Tensorflow_
- Version: 2.14.0
- Summary: TensorFlow is an open source machine learning framework for everyone.
- Home-page: https://www.tensorflow.org/
- Author: Google Inc.
- Author-email: [email protected]
- License: Apache 2.0
- Location: C:\Users\Sanju\AppData\Local\Programs\Python\Python311\Lib\site-packages
- Requires: tensorflow-intel
- _Keras_
- Name: keras
- Version: 2.14.0
- Summary: Deep learning for humans.
- Home-page: https://keras.io/
- Author: Keras team
- Author-email: [email protected]
- License: Apache 2.0
- Location: C:\Users\Sanju\AppData\Local\Programs\Python\Python311\Lib\site-packages
- Required-by: tensorflow-intel
Model Architecture:
Description:
Presenting a Recurrent Neural Network (RNN) model for Hindi OCR. Crafted for sequences, this model excels in recognizing patterns over time and space.
Key Features
- Adapted for Hindi character sequences.
- Incorporates SimpleRNN layers to process sequences.
- Enhanced performance metrics on the validation set.
- Changes from Previous Version:
Initial release. Model Performance Metrics: Training accuracy: 88% | Accuracy on unseen samples: 90%
Model Architecture:
The next milestone is to create an API that can use these models, and make it simple for method calling and classification.
@Khurdhula-Harshavardhan