app.py

import mediapipe as mp
import gradio as gr
import cv2
import torch


# Images
torch.hub.download_url_to_file('https://artbreeder.b-cdn.net/imgs/c789e54661bfb432c5522a36553f.jpeg', 'face1.jpg')
torch.hub.download_url_to_file('https://artbreeder.b-cdn.net/imgs/c86622e8cb58d490e35b01cb9996.jpeg', 'face2.jpg')

mp_face_mesh = mp.solutions.face_mesh

# Prepare DrawingSpec for drawing the face landmarks later.
mp_drawing = mp.solutions.drawing_utils 
drawing_spec = mp_drawing.DrawingSpec(thickness=1, circle_radius=1)

# Run MediaPipe Face Mesh.

def inference(image):
    with mp_face_mesh.FaceMesh(
        static_image_mode=True,
        max_num_faces=2,
        min_detection_confidence=0.5) as face_mesh:
        # Convert the BGR image to RGB and process it with MediaPipe Face Mesh.
        results = face_mesh.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))

        annotated_image = image.copy()
        for face_landmarks in results.multi_face_landmarks:
            mp_drawing.draw_landmarks(
                image=annotated_image,
                landmark_list=face_landmarks,
                connections=mp_face_mesh.FACE_CONNECTIONS,
                landmark_drawing_spec=drawing_spec,
                connection_drawing_spec=drawing_spec)
            return annotated_image

title = "Face Mesh"
description = "Gradio demo for Face Mesh. To use it, simply upload your image, or click one of the examples to load them. Read more at the links below."
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/1907.06724'>Real-time Facial Surface Geometry from Monocular Video on Mobile GPUs</a> | <a href='https://github.com/google/mediapipe'>Github Repo</a></p>"

gr.Interface(
    inference, 
    [gr.inputs.Image(label="Input")], 
    gr.outputs.Image(type="pil", label="Output"),
    title=title,
    description=description,
    article=article, 
    examples=[
              ["face1.jpg"],
              ["face2.jpg"]
    ]).launch()