This project implements a Convolutional Neural Network (CNN) to classify American Sign Language (ASL) hand gesture images into their respective alphabet classes (A-Z).
ASL is a vital communication method for the deaf and hard-of-hearing community. This project aims to automate the recognition of hand signs using deep learning.
The pipeline includes:
- Image Preprocessing: Grayscale conversion and normalization.
- Data Augmentation: Real-time transformations to prevent overfitting.
- CNN Training: Feature extraction using deep convolutional layers.
- Evaluation: Analyzing performance via accuracy/loss curves.
The model is built using a sequential architecture designed for spatial feature extraction:
- Convolutional Layers: 3 blocks of Conv2D for identifying edges and shapes.
- Batch Normalization: Applied after each conv layer for faster convergence.
- MaxPooling: Reduces the spatial dimensions of the data.
- Dropout (0.5): Ensures the model generalizes well to unseen images.
- Softmax Output: Classifies the input into one of the 26/29 alphabet categories.
- Language: Python
- Deep Learning: TensorFlow / Keras
- Data Science: NumPy, Matplotlib, Scikit-learn
- Computer Vision: OpenCV
Vansh Aggarwal B.Tech CSE Student
📧 Email Me: vansh27102005@gmail.com | 🔗 LinkedIn : https://www.linkedin.com/in/vanshaggarwal27/
This project is open-source and available under the MIT License.