utils1.py

import torch
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

cmap = sns.diverging_palette(262, 7, sep=1, n=16, s=99, l=50, center="dark", as_cmap=True) # best


def show(X):
	if X.dim() == 3 and X.size(0) == 3:
		plt.imshow( np.transpose(  X.numpy() , (1, 2, 0))  )
		plt.show()
	elif X.dim() == 2:
		plt.imshow(   X.numpy() , cmap='gray'  )
		plt.show()
	else:
		print('WRONG TENSOR SIZE')


def show_template(template):
	vm=0.3
	height=4
	width=6
	f, ax1 = plt.subplots()
	temp=template.detach().numpy()
	sns.heatmap(temp, cmap=cmap, ax=ax1, cbar=True,square=True,vmin=-vm, vmax=vm )
	plt.show()



def show_prob_mnist(p):

	p=p.data.squeeze().numpy()

	ft=15
	label = ('zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight','nine')
    # label = ('angry', 'disgust', 'fear', 'happy', 'neutral', 'sad', 'surprise')
	#p=p.data.squeeze().numpy()
	y_pos = np.arange(len(p))*1.2
	target=2
	width=0.9
	col= 'blue'
	#col='darkgreen'

	plt.rcdefaults()
	fig, ax = plt.subplots()

	# the plot
	ax.barh(y_pos, p, width , align='center', color=col)

	ax.set_xlim([0, 1.3])
	#ax.set_ylim([-0.8, len(p)*1.2-1+0.8])

	# y label
	ax.set_yticks(y_pos)
	ax.set_yticklabels(label, fontsize=ft)
	ax.invert_yaxis()
	#ax.set_xlabel('Performance')
	#ax.set_title('How fast do you want to go today?')

	# x label
	ax.set_xticklabels([])
	ax.set_xticks([])
	#x_pos=np.array([0, 0.25 , 0.5 , 0.75 , 1])
	#ax.set_xticks(x_pos)
	#ax.set_xticklabels( [0, 0.25 , 0.5 , 0.75 , 1] , fontsize=15)

	ax.spines['right'].set_visible(False)
	ax.spines['top'].set_visible(False)
	ax.spines['bottom'].set_visible(False)
	ax.spines['left'].set_linewidth(4)


	for i in range(len(p)):
	    str_nb="{0:.2f}".format(p[i])
	    ax.text( p[i] + 0.05 , y_pos[i] ,str_nb ,
	             horizontalalignment='left', verticalalignment='center',
	             transform=ax.transData, color= col,fontsize=ft)



	plt.show()
	#fig.savefig('pic/prob', dpi=96, bbox_inches="tight")





def show_prob_face(p):

	p = p.data.squeeze().numpy()

	ft=15
	label = ('angry', 'disgust', 'fear', 'happy', 'neutral', 'sad', 'surprise')
	#p=p.data.squeeze().numpy()
	y_pos = np.arange(len(p))*1.2
	target=2
	width=0.9
	col= 'green'
	#col='darkgreen'

	plt.rcdefaults()
	fig, ax = plt.subplots()

	# the plot
	ax.barh(y_pos, p, width , align='center', color=col)

	ax.set_xlim([0, 1.3])
	#ax.set_ylim([-0.8, len(p)*1.2-1+0.8])

	# y label
	ax.set_yticks(y_pos)
	ax.set_yticklabels(label, fontsize=ft)
	ax.invert_yaxis()
	#ax.set_xlabel('Performance')
	#ax.set_title('How fast do you want to go today?')

	# x label
	ax.set_xticklabels([])
	ax.set_xticks([])
	#x_pos=np.array([0, 0.25 , 0.5 , 0.75 , 1])
	#ax.set_xticks(x_pos)
	#ax.set_xticklabels( [0, 0.25 , 0.5 , 0.75 , 1] , fontsize=15)

	ax.spines['right'].set_visible(False)
	ax.spines['top'].set_visible(False)
	ax.spines['bottom'].set_visible(False)
	ax.spines['left'].set_linewidth(4)


	for i in range(len(p)):
	    str_nb="{0:.2f}".format(p[i])
	    ax.text( p[i] + 0.05 , y_pos[i] ,str_nb ,
	             horizontalalignment='left', verticalalignment='center',
	             transform=ax.transData, color= col,fontsize=ft)



	plt.show()
	#fig.savefig('pic/prob', dpi=96, bbox_inches="tight")


def show_prob_fashion_mnist(p):

	p=p.data.squeeze().numpy()

	ft=15
	label = ('T-shirt', 'Trouser', 'Pullover', 'Dress', 'Coat', 'Sandal', 'Shirt', 'Sneaker', 'Bag','Boot')
	#p=p.data.squeeze().numpy()
	y_pos = np.arange(len(p))*1.2
	target=2
	width=0.9
	col= 'blue'
	#col='darkgreen'

	plt.rcdefaults()
	fig, ax = plt.subplots()

	# the plot
	ax.barh(y_pos, p, width , align='center', color=col)

	ax.set_xlim([0, 1.3])
	#ax.set_ylim([-0.8, len(p)*1.2-1+0.8])

	# y label
	ax.set_yticks(y_pos)
	ax.set_yticklabels(label, fontsize=ft)
	ax.invert_yaxis()
	#ax.set_xlabel('Performance')
	#ax.set_title('How fast do you want to go today?')

	# x label
	ax.set_xticklabels([])
	ax.set_xticks([])
	#x_pos=np.array([0, 0.25 , 0.5 , 0.75 , 1])
	#ax.set_xticks(x_pos)
	#ax.set_xticklabels( [0, 0.25 , 0.5 , 0.75 , 1] , fontsize=15)

	ax.spines['right'].set_visible(False)
	ax.spines['top'].set_visible(False)
	ax.spines['bottom'].set_visible(False)
	ax.spines['left'].set_linewidth(4)


	for i in range(len(p)):
	    str_nb="{0:.2f}".format(p[i])
	    ax.text( p[i] + 0.05 , y_pos[i] ,str_nb ,
	             horizontalalignment='left', verticalalignment='center',
	             transform=ax.transData, color= col,fontsize=ft)



	plt.show()
	#fig.savefig('pic/prob', dpi=96, bbox_inches="tight")



import os.path
def check_mnist_dataset_exists(path_data=''):
    flag_train_data = os.path.isfile(path_data + 'mnist/train_data.pt')
    flag_train_label = os.path.isfile(path_data + 'mnist/train_label.pt')
    flag_test_data = os.path.isfile(path_data + 'mnist/test_data.pt')
    flag_test_label = os.path.isfile(path_data + 'mnist/test_label.pt')
    if flag_train_data==False or flag_train_label==False or flag_test_data==False or flag_test_label==False:
        print('MNIST dataset missing - downloading...')
        import torchvision
        import torchvision.transforms as transforms
        trainset = torchvision.datasets.MNIST(root=path_data + 'mnist/temp', train=True,
                                                download=True, transform=transforms.ToTensor())
        testset = torchvision.datasets.MNIST(root=path_data + 'mnist/temp', train=False,
                                               download=True, transform=transforms.ToTensor())
        train_data=torch.Tensor(60000,28,28)
        train_label=torch.LongTensor(60000)
        for idx , example in enumerate(trainset):
            train_data[idx]=example[0].squeeze()
            train_label[idx]=example[1]
        torch.save(train_data,path_data + 'mnist/train_data.pt')
        torch.save(train_label,path_data + 'mnist/train_label.pt')
        test_data=torch.Tensor(10000,28,28)
        test_label=torch.LongTensor(10000)
        for idx , example in enumerate(testset):
            test_data[idx]=example[0].squeeze()
            test_label[idx]=example[1]
        torch.save(test_data,path_data + 'mnist/test_data.pt')
        torch.save(test_label,path_data + 'mnist/test_label.pt')
    return path_data

def check_fashion_mnist_dataset_exists(path_data=''):
    flag_train_data = os.path.isfile(path_data + 'fashion-mnist/train_data.pt')
    flag_train_label = os.path.isfile(path_data + 'fashion-mnist/train_label.pt')
    flag_test_data = os.path.isfile(path_data + 'fashion-mnist/test_data.pt')
    flag_test_label = os.path.isfile(path_data + 'fashion-mnist/test_label.pt')
    if flag_train_data==False or flag_train_label==False or flag_test_data==False or flag_test_label==False:
        print('FASHION-MNIST dataset missing - downloading...')
        import torchvision
        import torchvision.transforms as transforms
        trainset = torchvision.datasets.FashionMNIST(root=path_data + 'fashion-mnist/temp', train=True,
                                                download=True, transform=transforms.ToTensor())
        testset = torchvision.datasets.FashionMNIST(root=path_data + 'fashion-mnist/temp', train=False,
                                               download=True, transform=transforms.ToTensor())
        train_data=torch.Tensor(60000,28,28)
        train_label=torch.LongTensor(60000)
        for idx , example in enumerate(trainset):
            train_data[idx]=example[0].squeeze()
            train_label[idx]=example[1]
        torch.save(train_data,path_data + 'fashion-mnist/train_data.pt')
        torch.save(train_label,path_data + 'fashion-mnist/train_label.pt')
        test_data=torch.Tensor(10000,28,28)
        test_label=torch.LongTensor(10000)
        for idx , example in enumerate(testset):
            test_data[idx]=example[0].squeeze()
            test_label[idx]=example[1]
        torch.save(test_data,path_data + 'fashion-mnist/test_data.pt')
        torch.save(test_label,path_data + 'fashion-mnist/test_label.pt')
    return path_data

def check_cifar_dataset_exists(path_data=''):
    flag_train_data = os.path.isfile(path_data + 'cifar/train_data.pt')
    flag_train_label = os.path.isfile(path_data + 'cifar/train_label.pt')
    flag_test_data = os.path.isfile(path_data + 'cifar/test_data.pt')
    flag_test_label = os.path.isfile(path_data + 'cifar/test_label.pt')
    if flag_train_data==False or flag_train_label==False or flag_test_data==False or flag_test_label==False:
        print('CIFAR dataset missing - downloading...')
        import torchvision
        import torchvision.transforms as transforms
        trainset = torchvision.datasets.CIFAR10(root=path_data + 'cifar/temp', train=True,
                                        download=True, transform=transforms.ToTensor())
        testset = torchvision.datasets.CIFAR10(root=path_data + 'cifar/temp', train=False,
                                       download=True, transform=transforms.ToTensor())
        train_data=torch.Tensor(50000,3,32,32)
        train_label=torch.LongTensor(50000)
        for idx , example in enumerate(trainset):
            train_data[idx]=example[0]
            train_label[idx]=example[1]
        torch.save(train_data,path_data + 'cifar/train_data.pt')
        torch.save(train_label,path_data + 'cifar/train_label.pt')
        test_data=torch.Tensor(10000,3,32,32)
        test_label=torch.LongTensor(10000)
        for idx , example in enumerate(testset):
            test_data[idx]=example[0]
            test_label[idx]=example[1]
        torch.save(test_data,path_data + 'cifar/test_data.pt')
        torch.save(test_label,path_data + 'cifar/test_label.pt')
    return path_data



def display_scores(sc):


    ft=10
    ft_label=12

    bs=sc.size(0)
    nb_class=sc.size(1)

    f, ax = plt.subplots(1, bs)

    if bs ==2:
         f.set_size_inches(8, 8)
         f.subplots_adjust(left=None, bottom=None, right=None, top=None,
                wspace=None, hspace=0.5)
    else:
         f.set_size_inches(12, 12)
         f.subplots_adjust(left=None, bottom=None, right=None, top=None,
                wspace=None, hspace=0.25)

    max_score= sc.max().item()
    min_score=sc.min().item()

    label_pos= min_score-8
    xmin=-5.5
    xmax=5.5


    label=[]
    for i in range(nb_class):
        str_nb="{0:.0f}".format(i)
        mystr='class '+ str_nb
        label.append(mystr)

    y_pos = np.arange(nb_class)*1.2



    for count in range(bs):

        ax[count].set_title('data point '+ "{0:.0f}".format(count))

        scores=sc[count].numpy()

        width=0.9
        col='darkgreen'

    #    plt.rcdefaults()

        # line in the middle
        ax[count].plot([0,0], [y_pos[0]-1,y_pos[-1]+1], color='k',linewidth=4)


        # the plot
        barlist=ax[count].barh(y_pos, scores, width , align='center', color=col)

        for idx,bar in enumerate(barlist):
            if scores[idx]<0:
                bar.set_color('r')

        ax[count].set_xlim([xmin, xmax])
        ax[count].invert_yaxis()

        # no y label
        ax[count].set_yticklabels([])
        ax[count].set_yticks([])

        # x label
        ax[count].set_xticklabels([])
        ax[count].set_xticks([])


        ax[count].spines['right'].set_visible(False)
        ax[count].spines['top'].set_visible(False)
        ax[count].spines['bottom'].set_visible(False)
        ax[count].spines['left'].set_visible(False)

        ax[count].set_aspect('equal')


        for i in range(len(scores)):

            str_nb="{0:.1f}".format(scores[i])
            if scores[i]>=0:
                ax[count].text( scores[i] + 0.05 , y_pos[i] ,str_nb ,
                     horizontalalignment='left', verticalalignment='center',
                     transform=ax[count].transData, color= col,fontsize=ft)
            else:
                ax[count].text( scores[i] - 0.05 , y_pos[i] ,str_nb ,
                     horizontalalignment='right', verticalalignment='center',
                     transform=ax[count].transData, color= 'r',fontsize=ft)

            if  count ==0:
                ax[0].text( label_pos , y_pos[i] , label[i] ,
                         horizontalalignment='left', verticalalignment='center',
                         transform=ax[0].transData, color= 'black',fontsize=ft_label)


    plt.show()


def num_correct_prediction( scores , labels ):

    predicted_labels = scores.argmax(dim=1)
    indicator = (predicted_labels == labels)
    num_matches=indicator.sum()
    num_matches_python= num_matches.item()

    return num_matches_python


def get_error( scores , labels ):

    bs=scores.size(0)
    predicted_labels = scores.argmax(dim=1)
    indicator = (predicted_labels == labels)
    num_matches=indicator.sum()

    return 1-num_matches.float()/bs


def show_prob_cifar(p):


    p=p.data.squeeze().numpy()

    ft=15
    label = ('airplane', 'automobile', 'Bird', 'Cat', 'Deer', 'Dog', 'Frog', 'Horse', 'Ship','Truck' )
    #p=p.data.squeeze().numpy()
    y_pos = np.arange(len(p))*1.2
    target=2
    width=0.9
    col= 'blue'
    #col='darkgreen'

    plt.rcdefaults()
    fig, ax = plt.subplots()

    # the plot
    ax.barh(y_pos, p, width , align='center', color=col)

    ax.set_xlim([0, 1.3])
    #ax.set_ylim([-0.8, len(p)*1.2-1+0.8])

    # y label
    ax.set_yticks(y_pos)
    ax.set_yticklabels(label, fontsize=ft)
    ax.invert_yaxis()
    #ax.set_xlabel('Performance')
    #ax.set_title('How fast do you want to go today?')

    # x label
    ax.set_xticklabels([])
    ax.set_xticks([])
    #x_pos=np.array([0, 0.25 , 0.5 , 0.75 , 1])
    #ax.set_xticks(x_pos)
    #ax.set_xticklabels( [0, 0.25 , 0.5 , 0.75 , 1] , fontsize=15)

    ax.spines['right'].set_visible(False)
    ax.spines['top'].set_visible(False)
    ax.spines['bottom'].set_visible(False)
    ax.spines['left'].set_linewidth(4)


    for i in range(len(p)):
        str_nb="{0:.2f}".format(p[i])
        ax.text( p[i] + 0.05 , y_pos[i] ,str_nb ,
                 horizontalalignment='left', verticalalignment='center',
                 transform=ax.transData, color= col,fontsize=ft)



    plt.show()


def display_num_param(net):
	nb_param = 0
	for param in net.parameters():
	    nb_param += param.numel()
	print('There are {} ({:.2f} million) parameters in this neural network'.format(
		nb_param, nb_param/1e6)
	     )