Facial Feature-Preserving Filters: Enhancing Facial Expressions in Filtered Videos

Project Description

This project aims to develop a script that applies visual filters to videos of human faces while preserving the core features of facial expressions. The primary goal is to maintain the subtle nuances of facial movements, even when applying transformative filters. This ensures that the original expressions remain recognizable and expressive.

Problem Statement

Traditional filters often distort or mask the fine details of facial expressions, which are crucial for conveying emotions. This project addresses the challenge of creating filters that maintain the integrity of these delicate movements, ensuring that the original expressions remain discernible.

Objectives and Goals

Develop a filter script that can be applied to videos, preserving facial expressions. Create at least three different filter options: Transforming the face into a cat Giving a more childlike appearance Providing a cartoonish look

Tools and Technologies

Python: The primary programming language used for development. MediaPipe: Specifically, the Face Mesh module is used for facial landmark detection and stylization. OpenCV: Used for image and video processing.

Data and Resources

Sample Videos: Provided for testing and demonstrating the filters developed. Image Files: Various face images used for creating cartoon filters.

Installation and Setup

git clone https://github.com/Shellygur4/project3.git

Navigate to the Project Directory:

cd C:\Users\shell\homweork\project\project3

Install Dependencies:

pip install -r requirements.txt

Ensure the MediaPipe Model (face_stylizer.task) is available at the specified path in the code.

Usage

For identification of cat person face images and adds a smile to them

place your input image in the project3\different filter useage\cat_person_mesh\data

run the image processing script :

project3\different filter useage\cat_person_mesh\main.py

recieve output image in the project3\different filter useage\cat_person_mesh\output

For chillike filter

place your input image in the project3\different filter useage\child_filter\data

run the image processing script : project3\different filter useage\Child_Filter\main.py

recieve output image in the project3\different filter useage\child_filter\output

for a cartoon filter to a single image:

place your input image in the project3\different filter useage\image_to_cartoon_image\data

Run the image processing script:

project3\different filter useage\image_to_cartoon_image\main.py

recieve output image in the project3\different filter useage\image_to_cartoon_image\output

For Video to cartoon frames

place your input video in the project3\different filter useage\video_to_cartoon_frames\data

Run the video processing script:

project3\different filter useage\video_to_cartoon_frames\main.py

recieve output images in the project3\different filter useage\video_to_cartoon_frames\output

For Video to cartoon video

place your input video in the project3\different filter useage\video_to_cartoon_video\data

Run the video processing script: project3\different filter useage\video_to_cartoon_video\main.py

recieve output video in the project3\different filter useage\video_to_cartoon_video\output

Expected Deliverables

GitHub Repository: The project's codebase.

Final Presentation: Output should include filtered videos\images.

Potential Impact

This project has significant implications for enhancing social interaction skills in children with autism. By creating filters that preserve facial expressions, the tool aims to help autistic children recognize and mimic facial expressions, potentially improving their social interaction skills.

Contributing

If you would like to contribute to this project, please fork the repository and submit a pull request with your changes. Ensure that all code adheres to the project's coding standards and includes appropriate test coverage.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For any questions or additional information, please contact:

Project Lead: Shelly Gur

Email: shellygur4@gmail.com

GitHub: Shellygur4