# Facial Emotion Recognition – SimpleCNN

A lightweight convolutional neural network (CNN) for **facial emotion recognition**.  
The model is trained to classify grayscale facial images into **7 emotion categories**.

## Model Details
- **Model type:** SimpleCNN (custom lightweight CNN)
- **Framework:** PyTorch
- **Task:** Image Classification
- **Input shape:** 1 × 48 × 48 (grayscale image)
- **Number of classes:** 7
- **Classes:**
  - 0 → Angry  
  - 1 → Disgust  
  - 2 → Fear  
  - 3 → Happy  
  - 4 → Sad  
  - 5 → Surprise  
  - 6 → Neutral  

## Intended Uses & Limitations
- ✅ Educational projects, demos, prototypes.  
- ❌ Not suitable for medical, psychological, or safety-critical applications.  
- ❌ May not generalize well outside datasets like FER2013.  

## How to Use
```python
from huggingface_hub import hf_hub_download
import torch
import json

from facial_emotion import SimpleCNN  # your model class

# Load config
config = json.load(open("config.json"))

# Build model
model = SimpleCNN(num_classes=config["num_classes"], in_channels=config["in_channels"])

# Load weights
checkpoint = hf_hub_download(repo_id="sreenathsree1578/facial_emotion", filename="pytorch_model.bin")
model.load_state_dict(torch.load(checkpoint, map_location="cpu"))
model.eval()

# Example inference
dummy = torch.randn(1, 1, 48, 48)  # dummy grayscale image
with torch.no_grad():
    out = model(dummy)
    pred = torch.argmax(out, dim=1).item()
    print("Predicted emotion:", config["labels"][str(pred)])