Claims:We Claim:
1. A System and method for the face identification of refugees:
2. A system and method for the face identification of refugees: as claimed in claim 1 involves following stages:
a) Collecting the Image Dataset
b) Splitting the Dataset into Training and Testing
c) Image Augmentation
d) Building the CNN
e) Full Connection
f) Training the Network
g) Testing
3. Collecting the Dataset In order to train our machine, we need a huge amount of data so that our model can learn from them by identifying out certain relations and common features related to the objects.
4. Splitting the dataset- To use the powers of the libraries, we first need to import them. After importing the libraries, we need to split our data into two parts- taining_set and test_set.
5. To increase the number of images we are using augmentation methods like rotation, zooming, flipping, shearing and shifting.
6. Building the CNN- This is most important step for our network. It consists of three parts -
1. Convolution
2. Polling
3. Flattening
7. A fully connected layer that takes the output of convolution/pooling and predicts the best label to describe the image
8. After augmentation, the model is trained by training data i.e. 80% of data collected is used to train the model. During training the model extracts the features in different convolution layers there are 13 convolution layers each with the filter size (2,2) and five maxpooling layers and each with size (2,2).
9. The model is evaluated for the remaining 20% of the dataset and performance of the model is calculated using various performance metrics.
, Description:The first step in the algorithm implementation is the conversion of RGB image then resizes it to (224,224) in dimension. Then to increase the number of images we are using augmentation methods like rotation, zooming, flipping, shearing and shifting.
During algorithm implementation each image in the dataset undergoes 18 rotations by rotating 10 degrees each time randomly.
After augmentation the model is trained by training data i.e. 80% of data collected is used to train the model. During training the model extracts the features in different convolution layers there are 13 convolution layers each with the filter size (2,2) and five maxpooling layers and each with size (2,2). And each neuron in convolution layer has Relu activation function attached to it.
Here the convolution layer summarizes the presence of features in the input image and maps the features. And the Relu layer is a activation function F(x ) = max (0,x), this function changes all the negative activations to 0.
Then the pooling layer operates upon each feature map separately to create new set of same numbers of pooled feature map in proposed method we have applied maxpooling it calculates maximum value for each patch of the feature map. Then there is dense layer followed by the dense layer we have softmax classifier in the output layer which classifies and predicts the family.
Graphs represent the loss and accuracy variation during training and validation. As we go on training the model the loss decreases and the Accuracy increases. Then once the training is complete after 5 epochs the model will be saved as an h5 file.