EmoteFlow Emotion Recognition CNN
A CNN model trained on FER2013 (and optionally fine-tuned on CK+) for real-time student emotion recognition. Part of the EmoteFlow learning engagement platform.
Labels
| Index | Emotion |
|---|---|
| 0 | Angry |
| 1 | Disgust |
| 2 | Fear |
| 3 | Happy |
| 4 | Sad |
| 5 | Surprise |
| 6 | Neutral |
Performance
- Test Accuracy: 0.4866
- Input: 48x48 grayscale image, normalized to [0, 1]
- Output: 7-class softmax probabilities
Usage
import numpy as np
import onnxruntime as ort
session = ort.InferenceSession("emoteflow_model.onnx")
image = np.random.rand(1, 48, 48, 1).astype(np.float32) # your preprocessed frame
result = session.run(None, {"input": image})
emotion_probs = result[0][0]
Training
- Phase 1: Trained on FER2013 (50 epochs max, early stopping)
- Phase 2: Fine-tuned on CK+ (if available, 20 epochs)
- Architecture: 4-block CNN (64→128→256→512) with BatchNorm + Dropout
- Downloads last month
- -