Global Journal of Computer Science and Technology, D: Neural & Artificial Intelligence, Volume 22 Issue 1

© 2022 Global Journals Neural Network Design using a Virtual Reality Platform Global Journal of Computer Science and Technology Volume XXII Issue I Version I 40 ( )D Year 2022 model.add(Flatten()) # adding first fully connected layer with 256 outputs model.add(Dense(numNeuronsFCL1, activation='relu')) #adding second fully connected layer 128 outputs model.add(Dense(numNueronsFCL2, activation='relu')) # adding softmax layer for the classification model.add(Dense(numClasses, activation='softmax')) Fig. 1 shows the network architecture graph, while figures 2 and 3 show, for example, the information relating to the selected layers Conv2D and Dense obtained by clicking directly on the chart. This information includes: the type of the layer; the name of the layer, whether the layer is trainable; what the data type is. For Convolutional layers: the number of filters; the kernel size; the strides; padding; data format and dilation rate. The activation function, whether bias is used, and how the kernels and (if applied) biases are initialized. Each layer has its own unique set of characteristics.