app.py

from sklearn.metrics.pairwise import cosine_similarity
from sentence_transformers import SentenceTransformer
import gradio as gr
import numpy as np


model = SentenceTransformer("clip-ViT-L-14")


def predict(im1, im2):
  embeddings = model.encode([im1, im2])
  embeddings_reshaped = [np.reshape(img, (1, -1)) for img in embeddings]
  sim = cosine_similarity(embeddings_reshaped[0], embeddings_reshaped[1])
  if sim > 0.72:
    return sim, "SAME PERSON, UNLOCK PHONE"
  else:
    return sim, "DIFFERENT PEOPLE, DON'T UNLOCK"


interface = gr.Interface(fn=predict,
                         inputs= [gr.Image(type="pil", source="webcam"),
                                  gr.Image(type="pil", source="webcam")],
                         outputs= [gr.Number(label="Similarity"),
                                   gr.Textbox(label="Message")]
                         )

interface.launch(debug=True)