app.py

from transformers import pipeline
import gradio as gr

# Initialiser la pipeline de classification d'image zero-shot avec le modèle CLIP spécifié
model_name = "openai/clip-vit-large-patch14-336"
classifier = pipeline("zero-shot-image-classification", model=model_name)


def classify_image(image, labels):
    image_to_classify = image

    labels = labels.split(",")

    labels_for_classification = labels
    scores = classifier(image_to_classify, candidate_labels=labels_for_classification)

    print(
        f"The highest score is {scores[0]['score']:.3f} for the label {scores[0]['label']}"
    )

    return f"The highest score is {scores[0]['score']:.3f} for the label {scores[0]['label']}"


# Créer l'interface Gradio
interface = gr.Interface(
    fn=classify_image,
    inputs=[
        gr.Image(label="Input Image Component", type="pil"),
        gr.Textbox("apple, banana, car, dog, elephant"),
    ],
    outputs="text",
    title="Zero shot image classification",
    description="classification: ",
)

# Lancer l'interface Gradio
if __name__ == "__main__":
    interface.launch(
        quiet=True,
    )