import torch

import torchvision

import torch.nn.functional as F

import pandas as pd

import numpy as np

class Block(torch.nn.Module):

    def __init__(self, in_channels, mid_channel, out_channels, max_pool_kernel_size, batch_norm=False):

        super().__init__()

        self.max_pool_kernel_size=max_pool_kernel_size

        self.conv1 = torch.nn.Conv2d(in_channels=in_channels, out_channels=mid_channel, kernel_size=3, padding=1)

        self.conv2 = torch.nn.Conv2d(in_channels=mid_channel, out_channels=out_channels, kernel_size=3, padding=1)

        self.batch_norm = batch_norm

        if batch_norm:

            self.bn1 = torch.nn.BatchNorm2d(mid_channel)

            self.bn2 = torch.nn.BatchNorm2d(out_channels)

            self.bn3 = torch.nn.BatchNorm2d(out_channels)

    def forward(self, x):

        x = self.conv1(x)

        if self.batch_norm:

            x = self.bn1(x)

        x = torch.nn.functional.relu(x, inplace=True)

        x = self.conv2(x)

        if self.batch_norm:

            x = self.bn2(x)

        x = torch.nn.functional.relu(x, inplace=True)

        if self.max_pool_kernel_size!=1:

            x = torch.nn.functional.max_pool2d(x, kernel_size=self.max_pool_kernel_size)

        if self.batch_norm:

            x = self.bn3(x)

        out = x

        return out

class CxlNet(torch.nn.Module):

    def up(self, x, size):

        return torch.nn.functional.interpolate(x, size=size, mode=self.upscale_mode)

    def down(self, x):

        return torch.nn.functional.max_pool2d(x, kernel_size=2)

    def __init__(self, in_channels, out_channels, batch_norm=False, upscale_mode="nearest",image_size=512):

        super().__init__()

        self.in_channels = in_channels

        self.out_channels = out_channels

        self.batch_norm = batch_norm

        self.upscale_mode = upscale_mode

        self.image_size=image_size

        self.enc1 = Block(in_channels, 32, 64,2, batch_norm)

        self.enc2 = Block(64, 64, 64, 2, batch_norm)

        self.enc3 = Block(64, 128, 128, 2, batch_norm)

        self.enc4 = Block(128, 256, 256, 2, batch_norm)

        #self.enc3 = Block(256, 128, 128, 2, batch_norm)

        #self.enc4 = Block(128, 64, 64, 2, batch_norm)

        self.dec3 = Block(512, 256, 256, 1, batch_norm)

        self.dec2 = Block(256, 128, 128, 1, batch_norm)

        self.dec1 = Block(128, 64, 64, 1, batch_norm)

        self.dec0 = Block(64, 32, out_channels, 1, batch_norm)

    def forward(self, x):

        enc1 = self.enc1(x)

        enc2 = self.enc2(enc1)

        enc3 = self.enc3(enc2)

        enc4 = self.enc4(enc3)

        outOfDec3 = self.dec3(torch.cat([enc1,

                                    self.up(enc2, enc1.size()[-2:]),

                                    self.up(enc3, enc1.size()[-2:]),

                                    self.up(enc4, enc1.size()[-2:]),

        ], 1))

        outOfDec2 = self.dec2(self.up(outOfDec3, (self.image_size,self.image_size)))

        outOfDec1 = self.dec1(outOfDec2)

        outOfDec0 = self.dec0(outOfDec1)

        return outOfDec0