You can deploy this model and start earning money today!

Tanneru/
Facial-Emotion-Detection-BEIT-Large

Name: Facial-Emotion-Detection-BEIT-Large
Availability: InStock

facial-expression-recognition

Views

No views yet

BEiT-Large for Emotion Detection on AffectNet

This model is fine-tuned from microsoft/beit-large-patch16-224-pt22k-ft22k for facial emotion recognition using a cleaned and balanced version of the AffectNet dataset.

🧠 Classes

The model predicts the following 7 basic emotion classes:

😠 anger
🤢 disgust
😨 fear
😀 happy
😐 neutral
😢 sad
😲 surprise

📊 Dataset Overview

Emotion	Train Samples	Test Samples
anger	1500	1718
disgust	1229	1248
fear	1512	1664
happy	2340	2704
neutral	2758	2368
sad	3091	1584
surprise	2119	1920

📈 Training Metrics

Epoch	Training Loss	Validation Loss	Accuracy
1	0.4552	0.5809	0.6917
2	0.3000	0.6669	0.7079
3	0.1473	0.7098	0.7378
4	0.0674	0.8904	0.7353
5	0.0291	0.9008	0.7452
6	0.0216	0.9844	0.7503
7	0.0118	1.0369	0.7522
8	0.0069	1.0992	0.7486
9	0.0035	1.0947	0.7482
10	0.0023	1.1336	0.7461

✅ Final Accuracy: ~74.6% on the test set

Training Configuration

The model was trained using the Hugging Face Trainer with the following main arguments:

num_train_epochs=10
per_device_train_batch_size=64
per_device_eval_batch_size=64
gradient_accumulation_steps=2
learning_rate=5e-5
fp16=True (mixed precision training)
eval_strategy="epoch"
save_strategy="epoch"
save_total_limit=2
load_best_model_at_end=True
metric_for_best_model="accuracy"

Confusion Matrix

🔧 How to Use

python

from transformers import BeitImageProcessor, BeitForImageClassification
from PIL import Image
import requests

image_path = '/RAF-DB/aligned/test_0031_aligned.jpg' # ⬅️ Replace with your image path
image = Image.open(image_path).convert("RGB")

processor = BeitImageProcessor.from_pretrained("Tanneru/Facial-Emotion-Detection-BEIT-Large")
model = BeitForImageClassification.from_pretrained("Tanneru/Facial-Emotion-Detection-BEIT-Large")

inputs = processor(images=image, return_tensors="pt")

outputs = model(**inputs)
logits = outputs.logits


predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", model.config.id2label[predicted_class_idx])

📄 License

This model is released under the Apache 2.0 License. You are free to use, modify, and distribute the model with attribution.

✍️ Author

Username: Tanneru
Base model: microsoft/beit-large-patch16-224-pt22k-ft22k

📚 Citation

If you use this model in your work, please cite:

bibtex

@misc{tanneru2025beit_affectnet,
  title={BEiT-Large fine-tuned on AffectNet for Emotion Detection},
  author={Tanneru},
  year={2025},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/Tanneru/Facial-Emotion-Detection-BEIT-Large}},
}

@article{bao2021beit,
  author       = {Hangbo Bao and Li Dong and Furu Wei},
  title        = {BEiT: BERT Pre-Training of Image Transformers},
  journal      = {CoRR},
  volume       = {abs/2106.08254},
  year         = {2021},
  url          = {https://arxiv.org/abs/2106.08254},
  archivePrefix = {arXiv},
  eprint       = {2106.08254},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}