xandrew94x / lis-detection

In this project has been implemented an algorithm to identify the italian sign language (LIS - Lingua Italiana dei Segni) through Opencv, Mediapipe Hand Detection and a K-nn classifier.

LIS-Detection

In this project has been implemented an algorithm to detect the italian sign language (LIS - Lingua Italiana dei Segni) through Opencv, Mediapipe Hand Detection and a K-nn classifier.

🔨 Technologies

• Python 3.8.13
• Mediapipe 0.9.2.1
• Opencv 4.6.0.66
• Sklearn 1.1.2
• Pickle 4.0

📚 Install

pip install - pip/requirements.txt

📊 Dataset

Dataset was acquired using a previously created tool: acqTool

• LIS alphabet letters were recorded (26 classes).
• 500 vectors for each classes,
• Vectors was normalized using the MinMaxScaler [0,1],
• Each vector is composed by 42 features:

$$V_{i} = (x_{1}, y_{1}, ... , x_{n}, y_{n})$$

where:

• and are the position in pixels,
• (, ) is nth landmark on 2D image. 8

Note: only right hand was used during the acquisition. Download Here

🚸 Classifier

The classifier chosen is K-NN ( sklearn ).

Confusion Matrix:

🚀 Launch

python hand_detection_main.py -c {classifier.pkl}

📄 License

This project is released under the GNU General Public License v3.0