app.py

import gradio as gr
from transformers import ViTFeatureExtractor, ViTForImageClassification
from PIL import Image
import requests
img1_url = 'https://static3.depositphotos.com/1003410/160/i/450/depositphotos_1607848-stock-photo-portrait-of-old-man.jpg'
img1 = Image.open(requests.get(img1_url, stream=True).raw)
img2_url = "https://img.freepik.com/free-photo/excited-screaming-young-woman-standing-isolated-yellow_176420-39645.jpg"
img2 = Image.open(requests.get(img2_url, stream=True).raw)
img3_url = "https://img.freepik.com/free-photo/cute-adorable-boy-studio_58702-7629.jpg"
img3 = Image.open(requests.get(img3_url, stream=True).raw)

def age_emot_classifier(input_image):

    # Init model, transforms
    model_age = ViTForImageClassification.from_pretrained('nateraw/vit-age-classifier')
    transforms_age = ViTFeatureExtractor.from_pretrained('nateraw/vit-age-classifier')

    model_emot = ViTForImageClassification.from_pretrained("yangswei/visual-emotion-classification")
    transforms_emot = ViTFeatureExtractor.from_pretrained("yangswei/visual-emotion-classification")

    # Transform our image and pass it through the model
    inputs_age = transforms_age(input_image, return_tensors='pt')
    output_age = model_age(**inputs_age)

    inputs_emot = transforms_emot(input_image, return_tensors='pt')
    output_emot = model_emot(**inputs_emot)

    # Predicted Class probabilities
    proba_age = output_age.logits.softmax(1)
    proba_emot = output_emot.logits.softmax(1)

    # Predicted Classes With Confidences
    labels_age = model_age.config.id2label
    confidences_age = {labels_age[i]: proba_age[0][i].item() for i in range(len(labels_age))}

    labels_emot = model_emot.config.id2label
    confidences_emot = {labels_emot[i]: proba_emot[0][i].item() for i in range(len(labels_emot))}

    return confidences_age, confidences_emot

output_age = gr.Label(num_top_classes=9, label="Age Prediction")
output_emotion = gr.Label(num_top_classes=8, label="Emotion Prediction")

with gr.Blocks(theme=gr.themes.Glass()) as demo:
    gr.Interface(fn=age_emot_classifier, inputs="image", outputs=[output_age, output_emotion],
                 examples=[img1, img2, img3])

demo.launch()