BRAIT-Machine-Learning

This Contains what Machine Learning's Team Do

The job of Machine Learning team is to learn and make machine learning model of Braille Translator. The following are the work stages of the Machine Learning team

1. Search Dataset

We collect the data from Kaggle and Google Images

Dataset Resource

• Kaggle
• Google Images

2. Data Preparation

After collecting the dataset, we do preparation data by cleaning the image that not suitable for our model, such as deleting and crop the image that still can be use

Link To Dataset

• Braille Dataset

3. Preprocessing Data

We use image augmentation to the original images to create additional training samples to expanding the training data with diverse variations and to reduce overfiting. Preprocess the data using Label Encoding imported from sklearn.preprocessing.

Link To Preprocessing File

• Preprocessing

• Distribution of The Dataset

• Dataset Split

We split the data into three folders. Namely train, validation and test.

4. Create Models and Training the Data

We created Convolutional Neural Network models using PyTorch Framework for Braille Classification

Link BRAIT Model File

• BRAIT Model

• Model Summary

• Train The Model

5. Evaluate The Model

After training the model, we carry out a model evaluation to ensure that the model that has been worked on can perform Braille translation tasks with a high level of accuracy.

• BRAIT Model Loss and Accuracy

6. Test The Model

Testing models in the BRAIT project have an important role in ensuring that the model can work effectively and reliably in Braille translation tasks.

Link To Testing File

• Testing

• BRAIT Model Testing

6. Save The Model

• Save Model

This is our saved model file - BRAIT_PYTORCH.pth