Layers | SIT.net model Description | |
Name | Description | |
Layer 1 | Image Input Layer | define input image sizes 512 × 512 |
Layer 2 | Transposed CNN | define filters size, number of filters, bias, strides for upsampling of image features to maintain the exact image information till the end of output images without any loss |
Layer 3 | Dropout Layer | Defines probability of dropping nodes to prevent overfitting |
Layer 4 | Batch Normalization Layer | Define mean and variance scale to make fast and stable analysis between layers |
Layer 5 | Flatten Layer | Helps neural network to learn more complex patterns and help te network for better prediction. |
Layer 6 | Gru Layer | Defines activation function and number of hidden units to check dependence between different time series data. |
Layer 7 | Fully Connected Layer | Define output size with the help of connection between every neurons of one layer to other layers to provide flexibility. |
Layer 8 | Softmax layer | Helps to convert scale of vector numbers into scale of vector probabilities for prediction |
Layer 9 | Classification Layer | To computer loss and accuracy using probability of above layer for analysis |