Face Emotion Recognition CNN (v2)

Model Description

Face Emotion Recognition CNN (v2) is a lightweight Convolutional Neural Network designed to classify facial expressions into seven basic emotion categories using grayscale facial images of size 48×48. The model is optimized for fast inference and practical deployment in real-time and resource-constrained environments.

The model was trained using the Adam optimizer and categorical cross-entropy loss, balancing performance and computational efficiency.

Intended Use

This model is intended for:

Facial emotion recognition from static images
Human–computer interaction systems
Emotion-aware applications
Academic research and educational projects

Out-of-Scope Use

The model should not be used for:

Surveillance, biometric identification, or user profiling
Medical, psychological, or legal decision-making
Any high-risk or safety-critical applications

Model Architecture

Architecture Type: Convolutional Neural Network (CNN)
Input: Grayscale facial images (48 × 48)
Output: Probability distribution over 7 emotion classes
Optimizer: Adam
Loss Function: Categorical Crossentropy

Training Configuration

Image Size: 48 × 48
Batch Size: 64
Epochs: 40
Learning Rate: 0.0005
Number of Classes: 7
Model Name: face_emotion_cnn_v2

Emotion Labels

Label ID	Emotion
0	Angry
1	Disgust
2	Fear
3	Happy
4	Neutral
5	Sad
6	Surprise

Dataset

The model was trained on a facial emotion dataset containing grayscale face images resized to 48×48 pixels. Images were preprocessed using normalization and face alignment where applicable. The dataset includes seven emotion categories.

Note: Dataset composition and demographic distribution may influence model bias and generalization.

Evaluation

The model was evaluated on a held-out test set with the following results:

Training Performance (Final Epoch)

Accuracy: 0.5976
Loss: 1.1217

Test Performance

Test Accuracy: 0.5942

These results reflect moderate performance, which is typical for compact CNN models trained on low-resolution facial emotion datasets.

Usage

Inference Example

from transformers import pipeline

classifier = pipeline(
    task="image-classification",
    model="your-username/face_emotion_cnn_v2"
)

predictions = classifier("face.jpg")
print(predictions)

Downloads last month: 15