app.py

import gradio as gr
from sentence_transformers import SentenceTransformer
from PIL import Image
from scipy.spatial import distance

model = SentenceTransformer('clip-ViT-B-32')

def image_distance(image1, image2) -> float:
    embedding1 = model.encode(sentences=[Image.fromarray(image1)], batch_size=128)
    embedding2 = model.encode(sentences=[Image.fromarray(image2)], batch_size=128)

    print(embedding1.size, embedding2.size)

    return 1 - distance.cosine(embedding1[0], embedding2[0])

demo = gr.Interface(fn=image_distance, inputs=["image", "image"], outputs="label")
demo.launch()