from keras.models import load_model from PIL import Image import numpy as np import cv2 #the following are to do with this interactive notebook code from matplotlib import pyplot as plt # this lets you draw inline pictures in the notebooks import pylab # this allows you to control figure size pylab.rcParams['figure.figsize'] = (10.0, 8.0) # this controls figure size in the notebook ###loading model### age_model = load_model('age_model_pretrained.h5') gender_model = load_model('gender_model_pretrained.h5') emotion_model = load_model('emotion_model_pretrained.h5') # Labels on Age, Gender and Emotion to be predicted age_ranges = ['1-5', '6-10', '11-15', '16-20', '21-25', '26-30', '31-35','36-40','41-50','51-60','61-70','71-80','81-90'] gender_ranges = ['male', 'female'] emotion_ranges= ['positive','negative','neutral'] ############################ import io import streamlit as st bytes_data=None img_file_buffer=st.camera_input("Take a picture") if img_file_buffer is not None: test_image = Image.open(img_file_buffer) st.image(test_image, use_column_width=True) st.write(type(test_image)) test_image = np.asarray(test_image) gray = cv2.cvtColor(test_image,cv2.COLOR_BGR2GRAY) face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml') faces = face_cascade.detectMultiScale(gray, 1.3, 5) i = 0 for (x,y,w,h) in faces: i = i+1 cv2.rectangle(test_image,(x,y),(x+w,y+h),(203,12,255),2) img_gray=gray[y:y+h,x:x+w] emotion_img = cv2.resize(img_gray, (48, 48), interpolation = cv2.INTER_AREA) emotion_image_array = np.array(emotion_img) emotion_input = np.expand_dims(emotion_image_array, axis=0) output_emotion= emotion_ranges[np.argmax(emotion_model.predict(emotion_input))] gender_img = cv2.resize(img_gray, (100, 100), interpolation = cv2.INTER_AREA) gender_image_array = np.array(gender_img) gender_input = np.expand_dims(gender_image_array, axis=0) output_gender=gender_ranges[np.argmax(gender_model.predict(gender_input))] age_image=cv2.resize(img_gray, (200, 200), interpolation = cv2.INTER_AREA) age_input = age_image.reshape(-1, 200, 200, 1) output_age = age_ranges[np.argmax(age_model.predict(age_input))] output_str = str(i) + ": "+ output_gender + ', '+ output_age + ', '+ output_emotion st.write(output_str) col = (0,255,0) cv2.putText(test_image, str(i),(x,y),cv2.FONT_HERSHEY_SIMPLEX,1,col,2) st.image(cv2.cvtColor(test_image, cv2.COLOR_BGR2RGB)) if bytes_data is None: st.stop()