The same paradigm is followed by the test data set. The first column of the dataset represents the class label of the image and the remaining 784 columns represent the 28 x 28 pixels. The dataset on Kaggle is available in the CSV format where training data has 27455 rows and 785 columns. These images belong to the 25 classes of English alphabet starting from A to Y ( No class labels for Z because of gesture motions ). This dataset contains 27455 training images and 7172 test images all with a shape of 28 x 28 pixels. In this article, we have used the American Sign Language (ASL) data set that is provided by MNIST and it is publicly available at Kaggle.

Finally, we will obtain the classification accuracy score of the CNN model in this task. We will evaluate the classification performance of our model using the non-normalized and normalized confusion matrices.

After successful training of the CNN model, the corresponding alphabet of a sign language symbol will be predicted. In this article, we will classify the sign language symbols using the Convolutional Neural Network (CNN). It can recognize the hand symbols and predict the correct corresponding alphabet through sign language classification. For deaf-mute people, computer vision can generate English alphabets based on the sign language symbols. It has also been applied in many support for physically challenged people.

Computer Vision has many interesting applications ranging from industrial applications to social applications.

Normalized classes discord plugin