from transformers import pipeline
import gradio as gr

# Load a pre-trained image classification model
model = pipeline("image-classification", model="google/vit-base-patch16-224")

# Define a function for detecting actions
def classify_image(image):
    predictions = model(image)
    return {pred["label"]: round(pred["score"], 4) for pred in predictions}

# Gradio interface
interface = gr.Interface(
    fn=classify_image,
    inputs=gr.Image(type="pil"),  # Accepts image input
    outputs="json",  # Outputs predictions
    title="Action Classifier",
    description="Upload an image, and the model will classify actions (e.g., standing, sitting)."
)

# Launch the app
if __name__ == "__main__":
    interface.launch(server_name="0.0.0.0")