app.py

import gradio as gr
from sklearn.metrics.pairwise import cosine_similarity
from sentence_transformers import SentenceTransformer
from PIL import Image
import cv2
import os
import numpy as np

def extract_face(im):
  
    prototxt_path = 'deploy.prototxt'
    caffemodel_path = 'weights.caffemodel'

    # Read the model
    cv2_model = cv2.dnn.readNetFromCaffe(prototxt_path, caffemodel_path) 

    #pil_image = PIL.Image.open('image.jpg')
    image = cv2.cvtColor(np.array(im), cv2.COLOR_RGB2BGR)
    #image = cv2.imread(im)

    (h, w) = image.shape[:2]
    blob = cv2.dnn.blobFromImage(cv2.resize(image, (300, 300)), 1.0, (300, 300), (104.0, 177.0, 123.0))

    cv2_model.setInput(blob)
    detections = cv2_model.forward()

    # Identify each face
    for i in range(0, detections.shape[2]):
      box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
      (startX, startY, endX, endY) = box.astype("int")

      confidence = detections[0, 0, i, 2]

      # If confidence > 0.5, save it as a separate file
      if (confidence > 0.5):
        frame = image[startY:endY, startX:endX]
        #PIL_image = Image.fromarray(frame)
        file_name = 'faces/' + str(np.random.randint(1,10)) + '_' + 'face.png'
        cv2.imwrite(file_name, frame)

    return file_name
    
def predict(im1, im2,thresh,model_name):

  if not isinstance(im1,str):
    im1_face = im1
    im2_face = im2

  else:

    im1_face = Image.open(im1)
    im2_face = Image.open(im2)
    
  model = load_model(model_name)

  sim=cosine_similarity(model.encode([im1_face,im2_face]))[0][1]

  if sim > thresh:
    return round(sim,2), "SAME PERSON, UNLOCK PHONE"
  else:
    return round(sim,2), "DIFFERENT PEOPLE, DON'T UNLOCK"

def load_model(model_name):

  model = SentenceTransformer(model_name)
  
  return model

title = """<h1 id="title">FaceID for Facial Recognition with Face Detector</h1>"""

models = ['clip-ViT-B-16','clip-ViT-B-32','clip-ViT-L-14']

twitter_link = """
[![](https://img.shields.io/twitter/follow/nickmuchi?label=@nickmuchi&style=social)](https://twitter.com/nickmuchi)
"""

css = '''
h1#title {
  text-align: center;
}
'''
demo = gr.Blocks(css=css)

with demo:
    gr.Markdown(title)
    gr.Markdown(twitter_link)
    model_options = gr.Dropdown(choices=models,label='Embedding Models',value=models[-1],show_label=True)
    thresh = gr.Slider(minimum=0.5,maximum=1,value=0.85,step=0.1,label='Confidence')
    
    with gr.Tabs():
        with gr.TabItem("Face ID with No Face Detection"):
            
            with gr.Row():
                with gr.Column():
                    nd_image_input_1 = gr.Image(label='Image 1',type='pil',source='webcam')
                    nd_image_input_2 = gr.Image(label='Image 2',type='pil',source='webcam')
                    
                with gr.Column():
                    sim = gr.Number(label="Similarity")
                    msg = gr.Textbox(label="Message")
                
            nd_but = gr.Button('Verify')
                            
        with gr.TabItem("Face ID with Face Detector"):
            
            with gr.Row():
                with gr.Column():
                    fd_image_1 = gr.Image(label='Image 1',type='pil',source='webcam')
                    fd_image_2 = gr.Image(label='Image 2',type='pil',source='webcam')

                with gr.Column():
                    face_1 = gr.Image(label='Face Detected 1',type='filepath')
                    face_2 = gr.Image(label='Face Detected 2',type='filepath')
                    fd_image_1.change(extract_face,fd_image_1,face_1)
                    fd_image_2.change(extract_face,fd_image_2,face_2)
      
                            
            with gr.Row():
                with gr.Column():
                    sim_1 = gr.Number(label="Similarity")
                    msg_1 = gr.Textbox(label="Message")
                                                    
            fd_but = gr.Button('Verify')          
                
            
    nd_but.click(predict,inputs=[nd_image_input_1,nd_image_input_2,thresh,model_options],outputs=[sim,msg],queue=True)
    fd_but.click(predict,inputs=[face_1,face_2,thresh,model_options],outputs=[sim_1,msg_1],queue=True)
    
    gr.Markdown("![visitor badge](https://visitor-badge.glitch.me/badge?page_id=nickmuchi-faceId-corise-project)")

demo.launch(debug=True,enable_queue=True)