import torch
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import numpy as np
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.data import Dataset


class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5, 2, 2)
        self.pool = nn.MaxPool2d(3, 2)
        self.conv2 = nn.Conv2d(6, 16, 5, 2, 2)
        self.fc1 = nn.Linear(784, 240)
        self.fc2 = nn.Linear(240, 84)
        self.fc3 = nn.Linear(84, 12)
        self.dropout = nn.Dropout(p=0.5)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, self.num_flat_features(x))
        x = self.dropout(x)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x

    def num_flat_features(self, x):
        size = x.size()[1:]  # all dimensions except the batch dimension
        num_features = 1
        for s in size:
            num_features *= s
        return num_features


class Memory(Dataset):
    def __init__(self, dataset_array, dataset_labels):
        self.labels = dataset_labels.astype(np.float64)
        self.images = dataset_array.astype(np.float64)
    def __len__(self):
        return self.images.shape[0]
    def __getitem__(self, idx):
        return self.images[idx], self.labels[idx]


def dataloader(batch_size, num_workers=0):
#    transform0 = transforms.Compose(
#    [transforms.Resize((128, 128)),
#     transforms.ToTensor()
#     ])

#    datasetmeanstd = torchvision.datasets.ImageFolder(root='TRUNK12', transform=transform0)
#    mean, std = calculate_img_stats_full(datasetmeanstd)
#    mean = mean.numpy()
#    std = std.numpy()
    mean, std = [(0.56337297, 0.5472399, 0.5224609), (0.13041796, 0.13301197, 0.139214)]

    train_transform = transforms.Compose(
    [transforms.Resize((128, 128)),
     transforms.ToTensor(),
     transforms.Normalize(mean, std)
     ])

    test_transform = transforms.Compose(
    [transforms.Resize((128, 128)),
     transforms.ToTensor(),
     transforms.Normalize(mean, std)
     ])

    print("Data loading...")
    trainset = torchvision.datasets.ImageFolder(root='train', transform=train_transform)
    trainset_length = trainset.__len__()
    trainloader = torch.utils.data.DataLoader(trainset, batch_size=trainset_length, shuffle=False, num_workers=num_workers, pin_memory=False)
    cvset = torchvision.datasets.ImageFolder(root='cv', transform=test_transform)
    cvset_length = cvset.__len__()
    cvloader = torch.utils.data.DataLoader(cvset, batch_size=cvset_length, shuffle=False, num_workers=num_workers, pin_memory=False)
    testset = torchvision.datasets.ImageFolder(root='test', transform=test_transform)
    testset_length = testset.__len__()
    testloader = torch.utils.data.DataLoader(testset, batch_size=testset_length, shuffle=False, num_workers=num_workers, pin_memory=False)
    testloader2 = torch.utils.data.DataLoader(testset, batch_size=4, shuffle=True, num_workers=num_workers, pin_memory=False)

    trainset_array = next(iter(trainloader))[0].numpy()
    trainset_labels = next(iter(trainloader))[1].numpy()
    trainset_m = Memory(trainset_array, trainset_labels)
    trainloader_m = torch.utils.data.DataLoader(trainset_m, batch_size=batch_size, shuffle=True, num_workers=num_workers, pin_memory=False)

    cvset_array = next(iter(cvloader))[0].numpy()
    cvset_labels = next(iter(cvloader))[1].numpy()
    cvset_m = Memory(cvset_array, cvset_labels)
    cvloader_m = torch.utils.data.DataLoader(cvset_m, batch_size=batch_size, shuffle=False, num_workers=num_workers, pin_memory=False)

    testset_array = next(iter(testloader))[0].numpy()
    testset_labels = next(iter(testloader))[1].numpy()
    testset_m = Memory(testset_array, testset_labels)
    testloader_m = torch.utils.data.DataLoader(testset_m, batch_size=batch_size, shuffle=False, num_workers=num_workers, pin_memory=False)
    print("Data loaded.")
    return trainloader_m, cvloader_m, testloader_m, testloader2


def calculate_img_stats_full(dataset):
    imgs_ = torch.stack([img for img,_ in dataset],dim=3)
    imgs_ = imgs_.view(3,-1)
    imgs_mean = imgs_.mean(dim=1)
    imgs_std = imgs_.std(dim=1)
    return imgs_mean, imgs_std


def training(trainloader, cvloader, testloader, net, opt, epochs, lr, weight_decay, step_size, gamma):
    criterion = nn.CrossEntropyLoss()
    adam = optim.Adam(net.parameters(), lr=lr, betas=(0.9, 0.999), eps=1e-08, weight_decay=weight_decay, amsgrad=False)
    adagrad = optim.Adagrad(net.parameters(), lr=lr, lr_decay=0, weight_decay=weight_decay, initial_accumulator_value=0)
    sgd = optim.SGD(net.parameters(), lr=lr, momentum=0.9, weight_decay=weight_decay)
    if opt == 1:
        optimizer = adam
    elif opt == 2:
        optimizer = adagrad
    else:
        optimizer = sgd
    scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=step_size, gamma=gamma)

    print("Training started...")
    for epoch in range(epochs):  # loop over the dataset multiple times

        running_loss = 0.0
        for i, data in enumerate(trainloader, 0):
            # get the inputs
            inputs, labels = data
            inputs = inputs.float().to(device)
            labels = labels.long().to(device)

            # zero the parameter gradients
            optimizer.zero_grad()

            # forward + backward + optimize
            outputs = net(inputs)
            loss = criterion(outputs, labels)
            loss.backward()
            optimizer.step()

            # print statistics
            running_loss += loss.item()

        print('Epoch [%d / %d], Loss: %.10f' % (epoch + 1, epochs, running_loss))

        if epoch % 20 == 19:
            print('Accuracy of the network on the train images: %d %%' % accshow(trainloader, net))
            print('Accuracy of the network on the cross-validation images: %d %%' % accshow(cvloader, net))
            print('Accuracy of the network on the test images: %d %%' % accshow(testloader, net))

        scheduler.step()
    print('Training finished.')


def checking(trainloader, cvloader, testloader, epochs, step_size, loops):
    cv_score, cv_train, cv_test, cv_opt, cv_rate, cv_decay, test_score, test_train, test_cv, test_opt, test_rate, test_decay = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
    opt_list = [1, 2, 3]
    rate_list = [0.001, 0.003, 0.005, 0.01, 0.03, 0.05]
    print("Checking started...")
    for l in rate_list:
        for w in rate_list:
            for o in opt_list:
                for i in range(loops):
                    net = Net()
                    net = net.to(device)
                    training(trainloader, cvloader, testloader, net, o, epochs, l, w, step_size, gamma=0.1)
                    train_acc = accshow(trainloader, net)
                    cv_acc = accshow(cvloader, net)
                    test_acc = accshow(testloader, net)
                    print("Train: %d %%. CV: %d %%. Test: %d %%. Opt: %d. Rate: %.3f. Decay: %.3f. Loop: %d" % (train_acc, cv_acc, test_acc, o, l, w, i+1))
                    if cv_acc > cv_score:
                        cv_score = cv_acc
                        cv_train = train_acc
                        cv_test = test_acc
                        cv_opt = o
                        cv_rate = l
                        cv_decay = w
                        cv_net = net
                    if test_acc > test_score:
                        test_score = test_acc
                        test_train = train_acc
                        test_cv = cv_acc
                        test_opt = o
                        test_rate = l
                        test_decay = w
                        test_net = net
    print("Checking finished.")
    print("Best cross-validation set accuracy:")
    print("Train: %d %%. CV: %d %%. Test: %d %%. Opt: %d. Rate: %.3f. Decay: %.3f." % (cv_train, cv_score, cv_test, cv_opt, cv_rate, cv_decay))
    print("Best test set accuracy:")
    print("Train: %d %%. CV: %d %%. Test: %d %%. Opt: %d. Rate: %.3f. Decay: %.3f." % (test_train, test_cv, test_score, test_opt, test_rate, test_decay))

    cv_net = cv_net.to(device)
    print("Checking best cross-validation set.")
    print('Accuracy of the network on the train images: %d %%' % accshow(trainloader, cv_net))
    print('Accuracy of the network on the cross-validation images: %d %%' % accshow(cvloader, cv_net))
    print('Accuracy of the network on the test images: %d %%' % accshow(testloader, cv_net))
    test_net = test_net.to(device)
    print("Checking best test set.")
    print('Accuracy of the network on the train images: %d %%' % accshow(trainloader, test_net))
    print('Accuracy of the network on the cross-validation images: %d %%' % accshow(cvloader, test_net))
    print('Accuracy of the network on the test images: %d %%' % accshow(testloader, test_net))


def accshow(loader, net):
    correct = 0
    total = 0
    with torch.no_grad():
        for data in loader:
            inputs, labels = data
            inputs = inputs.float().to(device)
            labels = labels.long().to(device)
            outputs = net(inputs)
            _, predicted = torch.max(outputs.data, 1)
            total += labels.size(0)
            correct += (predicted == labels).sum().item()
    return (100 * correct / total)


def classaccshow(loader, net, classes, classes_length):
    class_correct = list(0. for i in range(classes_length))
    class_total = list(0. for i in range(classes_length))
    with torch.no_grad():
        for data in loader:
            inputs, labels = data
            inputs = inputs.float().to(device)
            labels = labels.long().to(device)
            outputs = net(inputs)
            _, predicted = torch.max(outputs, 1)
            c = (predicted == labels).squeeze()
            try:
                for i in range(12):
                    label = labels[i]
                    class_correct[label] += c[i].item()
                    class_total[label] += 1
            except IndexError:
                pass
    for i in range(classes_length):
        print('Accuracy of %5s : %2d %%' % (classes[i], 100 * class_correct[i] / class_total[i]))


def imshow(img):
    img = img * 0.13 + 0.5     # unnormalize
    npimg = img.numpy()
    plt.imshow(np.transpose(npimg, (1, 2, 0)))
    plt.show()


def imgshow(loader, net, classes):
    dataiter = iter(loader)
    inputs, labels = dataiter.next()
    imshow(torchvision.utils.make_grid(inputs))
    print('GroundTruth: ', ' '.join('%5s' % classes[labels[j]] for j in range(4)))
    inputs = inputs.to(device)
    outputs = net(inputs)
    _, predicted = torch.max(outputs, 1)
    print('Predicted: ', ' '.join('%5s' % classes[predicted[j]] for j in range(4)))



if __name__ == '__main__':

    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

    classes = ('alder', 'beech', 'birch', 'chestnut', 'gingko biloba', 'hornbeam', 'horse chestnut', 'linden', 'oak', 'oriental plane', 'pine', 'spruce')
    classes_length = len(classes)
    batch_size = 60
    trainloader, cvloader, testloader, testloader2 = dataloader(batch_size)

    net = Net()
    net = net.to(device)
    epochs = 500
    step_size = 400
    opt = 1
    learning_rate = 0.003
    weight_decay = 0.03

    training(trainloader, cvloader, testloader, net, opt, epochs, learning_rate, weight_decay, step_size, gamma=0.1)

#    checking(trainloader, cvloader, testloader, epochs, step_size, loops=1)

#    classaccshow(cvloader, net, classes, classes_length)
#    classaccshow(testloader, net, classes, classes_length)
    imgshow(testloader2, net, classes)