app.py

'''
    Build Project
'''

# imported necessary libraries
import gradio as gr
import tensorflow as tf
import numpy as np
import os
import yaml
import cv2
from PIL import Image
import io
import json

# read config file
def read_config():
    config = {}
    print(os.path.curdir)
    with open('models_config.yaml', 'r') as cf:
        config = yaml.safe_load(cf)

    for var in config:
        config[var] = config[var].replace(';', os.sep)

    return config

config = read_config()

# loading the models
age_model = tf.keras.models.load_model(config['A_M_PATH'])
gender_model = tf.keras.models.load_model(config['G_M_PATH'])
face_cascade = cv2.CascadeClassifier(config['FD_M_PATH'])
 

def main_pipeline(image):
    
    #file = request.form()
    #im_b64 = file['img']

    #image = im_b64.file.read()

    image = Image.open(io.BytesIO(image))

    image = image.convert("RGB")

    image = np.asarray(image)

    gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
    faces = face_cascade.detectMultiScale(gray_img, 1.1, 4)
    
    if len(faces) == 0:
        return "NO FACE DETECTED"

    x, y, w, h =  faces[0]
    #cv2.rectangle(image, (x, y), (x+w, y+h), (255, 0, 0), 2)


    image = Image.fromarray(image)
    image = image.crop((x, y, x + w, y + h))

    #cv2.imshow('image', np.asarray(image))
    #cv2.waitKey()

    image = tf.image.resize(image, [224,224]) 
    image = tf.keras.preprocessing.image.img_to_array(image)
    image = image / 255.0      
    image = tf.expand_dims(image, axis=0)

    age_prds = age_model.predict(image)
    gender_prds = gender_model.predict(image)

    age_prds = np.around(age_prds)
    gender_prds = np.around(gender_prds)
    gender = ""

    if gender_prds[0][0] == 0:
        gender = 'male'
    else:
        gender = 'female'

    data = {}
    data['age'] = str(age_prds[0][0])
    data['gender'] = gender
    b_box_arr = [(x, y, w, h) for (x, y, w, h) in faces]

    data['left'] = str(b_box_arr[0][0])
    data['top'] =  str(b_box_arr[0][1])
    data['width'] = str(b_box_arr[0][2])
    data['height'] = str(b_box_arr[0][3])
    predictions_json = json.dumps(data)

    return predictions_json

# initializing the input component
image = gr.inputs.Image(shape = (299, 299, 3))
# initializing the output component
labels = gr.outputs.Label()

 
# launching the interface
gr.Interface(fn = main_pipeline, inputs = image,
             outputs = labels, capture_session = True).launch()