app.py

from transformers import ViTFeatureExtractor, ViTForImageClassification

import torch
import gradio as gr

from PIL import Image

feature_extractor = ViTFeatureExtractor.from_pretrained('google/vit-base-patch16-224')
model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224')

def classify_image(image):

    with torch.no_grad():
        model.eval()

        inputs = feature_extractor(images=image, return_tensors="pt")
        outputs = model(**inputs)
    
    logits = outputs.logits
    prob = torch.nn.functional.softmax(logits, dim=1)

    top10_prob, top10_indices = torch.topk(prob, 10)
    top10_confidences = {}
    for i in range(10):
        top10_confidences[model.config.id2label[int(top10_indices[0][i])]] = float(top10_prob[0][i])

    return top10_confidences #confidences


with gr.Blocks(title="ViT ImageNet Classification - ClassCat",
            css=".gradio-container {background:mintcream;}"
        ) as demo:
    gr.HTML("""<div style="font-family:'Times New Roman', 'Serif'; font-size:16pt; font-weight:bold; text-align:center; color:royalblue;">ViT - ImageNet Classification</div>""")

    with gr.Row(): 
        input_image = gr.Image(type="pil", image_mode="RGB", shape=(224, 224))
        output_label=gr.Label(label="Probabilities", num_top_classes=3)

    send_btn = gr.Button("Infer")
    send_btn.click(fn=classify_image, inputs=input_image, outputs=output_label)

#demo.queue(concurrency_count=3)
demo.launch(debug=True)