In this project, I utilized Convolutional Neural Networks (CNNs) to classify images of English Sign Language to their corresponding letters. I was inspired to take on this project when I stumbled upon a rich dataset on Kaggle and thought it would be both fun and challenging to work with.
To start, I preprocessed the dataset to ensure uniformity in the size and shape of the images.
Next, I split the dataset into training, validation, and test sets. For the model architecture, I experimented with different layers, filters, and kernel sizes to find the optimal combination that would yield the highest accuracy.
After training the CNN model on the training set, I evaluated its performance on the validation set and fine-tuned the hyperparameters to improve accuracy. Finally, I tested the model on the test set and obtained a high accuracy score.
This deep learning project provided me with a great opportunity to learn and experiment with CNNs for image classification. Through this project, I developed a better understanding of how to preprocess image data, optimize model architecture, and fine-tune hyperparameters for optimal performance.