Image classification for Tamil Alphabets using AI
MINST dataset is called the “Hello World” of Image classification. In this project, we will use the Tamil Alphabet for image classification. TensorFlow tutorials have an image classifier example used to classify flowers. We will take the same model to train Tamil Alphabets.
In this project, you will learn
- How to prepare a dataset
- How to retrain your dataset on a pre-defined architecture
- How to validate the generated model
Plan of Attack:
The following are the steps followed for image classification of Tamil alphabet
Step 1: Download dataset (We are using dataset collected by HP India).
Step 2: Make folders numbering from 1 to 155 to place the images.
Step 3: Read the downloaded dataset and sort the pictures based on names. For example, images start with the title “1” are images of the same alphabet.
Step 4: The dataset images are in .tiff format we need to convert them to .jpeg format for training.
Step 5: There are redundant images in two formats .tiff and .jpeg. In this step, we will delete the .tiff files.
Step 6: Run the retrain.py file on the dataset created to train and generate a model.
Step 7: Run label_image.py file to validate the generated model.
