Dc-4nderson's picture
Update README.md
1e38540 verified
metadata
license: mit
datasets:
  - Dc-4nderson/feelings_classfication_dataset
language:
  - en
base_model:
  - google/vit-base-patch16-224-in21k
pipeline_tag: image-classification
library_name: transformers
tags:
  - emotions
  - school
metrics:
  - accuracy

🤖 ViT Emotion Classifier

This is a lightweight Vision Transformer (ViT) model fine-tuned to classify emotions from facial images using a custom dataset of school-aged individuals. It supports 8 emotional categories and is designed to work well on small datasets and limited compute.


🧠 Supported Emotions

The model predicts one of the following emotional states:

Label ID Emotion
0 anxious-fearful
1 bored
2 confused
3 discouraged
4 frustrated
5 neutral
6 positive
7 suprised

📦 Model Details

  • Model Type: ViTForImageClassification
  • Backbone: vit-small-patch16-224
  • Dataset: Dc-4nderson/feelings_classfication_dataset
  • Framework: PyTorch
  • Labels: 8 emotions (defined in config.json)
  • Trained on: Google Colab with < 600 images

🧪 Usage

from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import torch

# Load model + processor
processor = AutoImageProcessor.from_pretrained("Dc-4nderson/vit-emotion-classifier")
model = AutoModelForImageClassification.from_pretrained("Dc-4nderson/vit-emotion-classifier")

# Load image and preprocess
image = Image.open("your_image.jpg").convert("RGB")
inputs = processor(images=image, return_tensors="pt")

# Run inference
with torch.no_grad():
    outputs = model(**inputs)
    pred = torch.argmax(outputs.logits, dim=1).item()
    label = model.config.id2label[str(pred)]

print("🧠 Predicted Emotion:", label)