This repository contains code implementing a lips state detection and classifier model powered by mediapipe - google's state of the art CNN and was inspired by the techniques outlined in our paper titled "Modelling Lips State Detection Using CNN for Non-verbal Communications"
The primary goal of this project is to detect and classify different states of the lips using advanced computer vision techniques and deep learning. The model leverages the capabilities of MediaPipe for robust facial landmark detection and incorporates the landmark information into a GridSearchCV optimized Support Vector Classifier for accurate lips state classification. The proposed solution delineates solutions that can be applied in non-verbal communication analysis.
Accurate detection of lip features through facial landmark analysis using MediaPipe.
Includes functions for realtime application and classification.
A grid search optimized Support Vector Classifier Model for classifying different lips states with examples of training and applications.
Our approach is grounded in research, as detailed in the paper "Modelling Lips State Detection Using CNN for Non-verbal Communications."
Make sure to install the required dependencies before running the code. You can find them in the requirements.txt file.
conda create --name --file requirements.txt
Clone the repository: git clone https://github.com/abtahiishmam3/Modelling-Lips-State-Detection-Mediapipe.git
Navigate to the project directory: cd Modelling-Lips-State-Detection-Mediapipe
Run the main notebook: Modelling-Lips-State-Detection-Mediapipe.ipynb
If you find this work useful in your research, please consider citing our paper.
Ishmam, A., Hasan, M., Hassan Onim, M.S., Roy, K., Hoque Akif, M.A., Nyeem, H. (2022). Modelling Lips State Detection Using CNN for Non-verbal Communications. In: Hossain, S., Hossain, M.S., Kaiser, M.S., Majumder, S.P., Ray, K. (eds) Proceedings of International Conference on Fourth Industrial Revolution and Beyond 2021 . Lecture Notes in Networks and Systems, vol 437. Springer, Singapore. https://doi.org/10.1007/978-981-19-2445-3_5
We welcome contributions from the community! If you would like to contribute to this project, please follow our contribution guidelines.
This project is licensed under the MIT License - see the LICENSE file for details.