Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/apoorva-info/eye-blink-detetction
Blink Detector Sensitive to Facial Orientation
https://github.com/apoorva-info/eye-blink-detetction
mediapipe numpy opencv python3
Last synced: 1 day ago
JSON representation
Blink Detector Sensitive to Facial Orientation
- Host: GitHub
- URL: https://github.com/apoorva-info/eye-blink-detetction
- Owner: apoorva-info
- Created: 2024-06-25T13:29:09.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2024-08-31T14:05:01.000Z (5 months ago)
- Last Synced: 2024-11-30T19:11:34.748Z (2 months ago)
- Topics: mediapipe, numpy, opencv, python3
- Language: Python
- Homepage:
- Size: 65.4 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Blink Detector Sensitive to Facial Orientation
## Overview
This project implements a blink detector that is sensitive to facial orientation. The program uses a webcam to capture video, processes the video to detect blinks, and records various metrics in a CSV file. The blink detection is influenced by the orientation of the face, adjusting the detection criteria based on whether the face is looking forward, left, right, or downward.
## Installation
1. Clone the repository:
```sh
git clone https://github.com/ansh-info/Eye-Blink-Detetction.git
```
2. Navigate to the project directory:
```sh
cd Eye Blink Detetction
```
3. Install the required packages:
```sh
pip install -r requirements.txt
```
### Additional Packages
- math (standard library)
- csv (standard library)
- time (standard library)
- datetime (standard library)
- sys (standard library)
### Hardware
- A functional webcam
## Usage
To use the blink detector, execute the program via the shell or terminal:
```bash
python blink_detector_face_orientation_datetime.py test_data
```
`test_data` is a placeholder for the participant ID. If none is provided, it defaults to `test`.
The program will open a video stream from your webcam, displaying the facial orientation and the number of detected blinks. To exit the program, press `ESC`.
### CSV Output
A CSV file will be created in the background containing the following columns:
- `time`: Time in seconds since the program started
- `blink`: Boolean, `True` if a blink is detected
- `ratio`: Average openness ratio of both eyes
- `ratio_r`: Openness ratio of the right eye
- `ratio_l`: Openness ratio of the left eye
- `counted`: Number of detected blinks
- `orientation`: Facial orientation (`l` for left, `r` for right, `d` for down, `f` for forward)
- `still_closed`: Boolean, `True` if the eyes remain closed (e.g., longer blink)
- `blink_length`: Time difference in seconds to the last measurement, accumulates if `still_closed` is `True`
### Blink Detection Based on Orientation
- Forward or Downward: Uses the average ratio of both eyes.
- Left: Uses the ratio of the right eye.
- Right: Uses the ratio of the left eye.
## Note
The program is sensitive to camera positioning. It works best when the camera is at the same height as the eyes. Higher camera angles still work well for forward orientation, but may incorrectly count blinks when looking left or right. Downward orientation is also more challenging for accurate blink detection.
The program performs better if the subject does not wear glasses.
## References
- [Eye Blink Detector and Counter with Mediapipe](https://medium.com/@aiphile/eyes-blink-detector-and-counter-mediapipe-a66254eb002c)
- [Head Pose Estimation using Python](https://towardsdatascience.com/head-pose-estimation-using-python-d165d3541600)