vit-face-affectnet

ViT Face Expression (Fine-tuned on AffectNet)

This model is a fine-tuned version of trpakov/vit-face-expression on the AffectNet dataset.

Model Description

• Architecture: Vision Transformer (ViT)
• Task: Facial Emotion Recognition
• Emotions: Anger, Disgust, Fear, Happiness, Neutral, Sadness, Surprise

Dataset

AffectNet is a large-scale database of facial expressions in the wild, containing more than 1M facial images from the Internet. This model was fine-tuned on a subset of the manually annotated images covering 7 basic emotions (excluding Contempt to align with the base model's taxonomy).

Usage

from transformers import ViTImageProcessor, ViTForImageClassification
from PIL import Image
import requests

url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)

repo_name = "michaelgathara/vit-face-affectnet"

processor = ViTImageProcessor.from_pretrained(repo_name)
model = ViTForImageClassification.from_pretrained(repo_name)

inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)
logits = outputs.logits
# model predicts one of the 7 emotions
predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", model.config.id2label[predicted_class_idx])