Facial-Emotion-Recognition is a project that focuses on detecting human emotions from facial expressions. It utilizes artificial intelligence (AI) to recognize emotions by analyzing facial features and expressions. The goal is to develop a system that can accurately interpret and respond to human emotions, similar to how the human brain does.
- About
- Eye Blinker Counter
- Facial Expression Recognition using custom CNN
- Smile Percentage Detection
- Face Emotions Recognition using Deep Learning
Facial-Emotion-Recognition is a comprehensive project that includes various components for emotion detection and recognition from facial expressions. The key components of this project are described below.
The Eye Blinker Counter component is responsible for detecting blinks in the eyes, which can be indicative of certain emotions. It utilizes facial landmark detection to locate the eyes in a video stream. The eye aspect ratio is calculated based on the eye coordinates, allowing the system to determine if a person is blinking. This information can be used to infer their emotional state.
- The first step is facial landmark detection to localize the eyes.
- Each eye is represented by coordinates, and the eye aspect ratio is calculated.
- A person is considered to be blinking if the eye aspect ratio rapidly approaches zero.
For more details, refer to the Eye Blinker Counter documentation.
Facial Expression Recognition using custom CNN is a model that utilizes Convolutional Neural Networks (CNNs) to analyze facial images and recognize different facial expressions. The model consists of two levels: background removal and expressional vector extraction.
- The first level involves background removal to extract emotions from an image.
- The second level uses a CNN to extract the primary expressional vector (EV) by tracking relevant facial points.
- The EV represents changes in expression and is used for emotion recognition.
For more details, refer to the Facial Expression Recognition documentation.
The Smile Percentage Detection component focuses on detecting smiles in real-time. It utilizes optical flow and facial feature tracking to determine if a person is smiling.
- The first human face is detected in the initial image frame, and standard facial features are located.
- Optical flow is used to track the positions of the left and right mouth corners in subsequent frames.
- If the distance between the tracked mouth corners exceeds a threshold, a smile is detected.
For more details, refer to the Smile Percentage Detection documentation.
Face Emotions Recognition using Deep Learning aims to classify human facial emotions using a Deep Convolutional Neural Network (DCNN) model. The system can process real-time facial images captured through a front camera and predict emotions.
- The DCNN model consists of convolution layers, pooling layers, and the sigmoid activation function.
- It is trained on a labeled facial image dataset to classify emotions into different classes.
- The model's performance is evaluated using a public database, achieving an accuracy of approximately 65%.
For more details, refer to the Face Emotions Recognition documentation.
| Component | Specifications |
|---|---|
| Eye Blinker Counter | - Facial landmark detection for eye localization |
| - Calculation of eye aspect ratio for blink detection | |
| Facial Expression Recognition | - Two-level CNN framework |
| Using custom CNN | - Background removal for emotion extraction |
| - Tracking of facial points for expressional vector | |
| Smile Percentage Detection | - Optical flow for mouth corner tracking |
| - Threshold-based smile detection | |
| Face Emotions Recognition | - Deep Convolutional Neural Network (DCNN) model |
| Using Deep Learning | - Convolution layers, pooling layers, sigmoid activation |
| - Trained on labeled facial image dataset | |
| - Achieves an accuracy of approximately 65% |
- Eye Blinker Counter documentation
- Facial Expression Recognition documentation
- Smile Percentage Detection documentation
- Face Emotions Recognition documentation
Please refer to the respective documentation for more detailed information about each component.