warriorwizard / Face_Age_Gender_Detection_Functional_CNN_Model

uses functional model api of keras to build age and gender detection from scratch

Face_Age_Gender_Detection_Functional_CNN_Model_using_Transfer_Learning

uses functional model api of keras to build age and gender detection from scratch

Using Transfer learning to train the model

from tensorflow.keras.models import Model
from tensorflow.keras.applications import ResNet50
from tensorflow.keras.layers import *
resnet = ResNet50(include_top= False, weights= 'imagenet',input_shape = (200,200,3))
resnet.trainable=  False
output = resnet.layers[-1].output

flatten = Flatten()(output)

dense1 = Dense(512, activation='relu')(flatten)
dense2 = Dense(512,activation='relu')(flatten)

dense3 = Dense(512,activation='relu')(dense1)
dense4 = Dense(512,activation='relu')(dense2)

output1 = Dense(1,activation='linear',name='age')(dense3)
output2 = Dense(1,activation='sigmoid',name='gender')(dense4)
model = Model(inputs=resnet.input,outputs=[output1,output2])
# Early stopping and Reduce learning rate if the model is not improving
from tensorflow.keras.callbacks import EarlyStopping
from tensorflow.keras.callbacks import ReduceLROnPlateau

earlystop = EarlyStopping(monitor='val_loss', patience=10, verbose=1, restore_best_weights=True)
learning_rate_reduction = ReduceLROnPlateau(monitor='val_loss', patience=5, verbose=1, factor=0.2, min_lr=0.00001)
model.compile(optimizer = 'Adam', loss = {'age': 'mae', 'gender': 'binary_crossentropy'},  metrics={'age': 'mae', 'gender': 'accuracy'},loss_weights={'age':1,'gender':99})