# deep learning imports
import torch
from torch import nn
import torch.nn.functional as fun


# Traffic Sign Classifier


class TrafficSignClassifier(nn.Module):
    def __init__(self, number_classes=43):
        super().__init__()

        # CNN layers
        self.conv1 = nn.Conv2d(3, 100, kernel_size=5)
        self.bn1 = nn.BatchNorm2d(100)
        self.conv2 = nn.Conv2d(100, 150, kernel_size=3)
        self.bn2 = nn.BatchNorm2d(150)
        self.conv3 = nn.Conv2d(150, 250, kernel_size=3)
        self.bn3 = nn.BatchNorm2d(250)
        self.conv_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(250 * 2 * 2, 350)
        self.fc2 = nn.Linear(350, number_classes)

        self.localization = nn.Sequential(
            nn.Conv2d(3, 8, kernel_size=7),
            nn.MaxPool2d(2, stride=2),
            nn.ReLU(True),
            nn.Conv2d(8, 10, kernel_size=5),
            nn.MaxPool2d(2, stride=2),
            nn.ReLU(True),
        )

        # Regressor for the 3 * 2 affine matrix
        self.fc_loc = nn.Sequential(
            nn.Linear(10 * 4 * 4, 32), nn.ReLU(True), nn.Linear(32, 3 * 2)
        )

        # Initialize the weights/bias with identity transformation
        self.fc_loc[2].weight.data.zero_()
        self.fc_loc[2].bias.data.copy_(
            torch.tensor([1, 0, 0, 0, 1, 0], dtype=torch.float)
        )

    # Spatial transformer network forward function
    def stn(self, x):
        xs = self.localization(x)
        xs = xs.view(-1, 10 * 4 * 4)
        theta = self.fc_loc(xs)
        theta = theta.view(-1, 2, 3)
        grid = fun.affine_grid(theta, x.size(), align_corners=True)
        x = fun.grid_sample(x, grid, align_corners=True)
        return x

    def forward(self, x):
        # transform the input
        x = fun.interpolate(x, size=(32, 32), mode="bilinear")
        x = self.stn(x)

        # Perform forward pass
        x = self.bn1(fun.max_pool2d(fun.leaky_relu(self.conv1(x)), 2))
        x = self.conv_drop(x)
        x = self.bn2(fun.max_pool2d(fun.leaky_relu(self.conv2(x)), 2))
        x = self.conv_drop(x)
        x = self.bn3(fun.max_pool2d(fun.leaky_relu(self.conv3(x)), 2))
        x = self.conv_drop(x)
        x = x.view(-1, 250 * 2 * 2)
        x = fun.relu(self.fc1(x))
        x = fun.dropout(x, training=self.training)
        x = self.fc2(x)
        return fun.log_softmax(x, dim=1)