✋ Sign Language Recognition using CNN

This project implements an American Sign Language (ASL) recognition system using a Convolutional Neural Network (CNN).
It predicts hand gestures representing alphabets and supports real-time webcam recognition.

---

📂 Dataset

• Source: ASL Alphabet Dataset - Kaggle
• Contains 87,000+ images of hand signs across 29 classes (A–Z, SPACE, DELETE, NOTHING).
• Images resized to 64×64 RGB before training.

---

🧠 Model

• Architecture: Custom CNN (Convolutional Neural Network)
• Input size: 64 × 64 × 3
• Output classes: 29 (A–Z + special tokens)
• Optimizer: Adam
• Loss Function: Categorical Crossentropy

The trained model is saved as SLR_final.h5.

⚠️ The model file is too large for GitHub (>100 MB).
👉 Download it here: Google Drive Link

---

📊 Results

✅ Accuracy

• Training Accuracy: 98.86%
• Validation Accuracy: 95.89%
• Test Accuracy (Augmented): 95.29%

✅ Confusion Matrix

The confusion matrix shows the model’s performance across different classes:

---

🎯 Sample Predictions

Here are some example outputs from the trained model:

Sign	Prediction
	A
	B
	C
	D
	E
	F

---

⚙️ Installation & Usage

1. Clone the Repository

git clone https://github.com/KrishSharma1903/Sign-Language-Recognition.git
cd Sign-Language-Recognition

2. Install Dependencies

pip install -r requirements.txt

3. Download Pretrained Model

Download SLR_final.h5 from Google Drive
and place it in the project root.

4. Run Notebook

jupyter notebook SignLanguageRecognition.ipynb

5. Live Prediction via Webcam

python model_load.py

---

🏗 Project Structure

Sign-Language-Recognition/
├── Results/
│   ├── A.png
│   ├── B.png
│   ├── C.png
│   ├── D.png
│   ├── E.png
│   ├── F.png
│   └── Confusion_Matrix.png
├── model_load.py
├── test_Data_Split.py
├── requirements.txt
├── SignLanguageRecognition.ipynb
└── .gitignore

---

🏆 Features

• ASL alphabet recognition (A–Z + SPACE, DELETE, NOTHING)
• Custom CNN architecture
• Confusion matrix + classification report
• Real-time webcam prediction with OpenCV
• Dataset splitting script (test_Data_Split.py)

---

👨‍💻 Authors

• Krish Sharma

---

📜 License

This project is open-source and available under the MIT License.