🙂 CNN — Facial Expression Recognition
CNN-based facial expression classifier trained to recognize 7 emotion categories from face images with a clean, reproducible pipeline.
📌 Model Summary
| Property | Details |
|---|---|
| 🏗️ Architecture | CNN (custom) |
| 🎯 Task | Image Classification |
| 😀 Classes | 7 emotions |
| ⚙️ Framework | PyTorch |
| 📐 Input size | 48 × 48 px (grayscale) |
| 📜 License | MIT |
🧠 Emotion Classes
| # | Emotion |
|---|---|
| 0 | 😠 Angry |
| 1 | 🤢 Disgust |
| 2 | 😨 Fear |
| 3 | 😄 Happy |
| 4 | 😐 Neutral |
| 5 | 😢 Sad |
| 6 | 😲 Surprise |
🚀 How to Use
import torch
import torch.nn.functional as F
from torchvision import transforms
from PIL import Image
# Load model
model = torch.jit.load("model.pt", map_location="cpu")
model.eval()
# Preprocessing
transform = transforms.Compose([
transforms.Grayscale(),
transforms.Resize((48, 48)),
transforms.ToTensor(),
transforms.Normalize([0.5], [0.5]),
])
EMOTIONS = ["Angry", "Disgust", "Fear", "Happy", "Neutral", "Sad", "Surprise"]
# Inference
image = Image.open("face.jpg")
tensor = transform(image).unsqueeze(0)
with torch.no_grad():
probs = F.softmax(model(tensor), dim=1)[0]
predicted = EMOTIONS[probs.argmax()]
confidence = probs.max().item()
print(f"Prediction: {predicted} ({confidence:.0%})")
🗂️ Training Data
- Base dataset: FER-2013 (Facial Expression Recognition)
- Input format: 48×48 grayscale face images
- Classes: 7 universal emotion categories
⚠️ Limitations
- Optimized for frontal face images
- Performance may degrade with partial occlusion, extreme lighting, or non-frontal poses
- Not intended for surveillance or identity recognition — expression classification only
🔗 Related Resources
👤 Author
Martin Badrous — Computer Vision & Deep Learning Engineer
