📑 Exercises with Drawing Images

🌐   DL_PP8_2_SMC.html

📑 Deep Learning. Practice Project 8_2:
Exercises with Drawing Images

✒️ Code Modules, Helpful Functions, & Settings

xxxxxxxxxx

 

1

#!python3 -m pip install tensorflow==2.6.0 \

2

#--user --quiet --no-warn-script-location

3

!python3 -m pip install tensorflow_hub \

4

--user --quiet --no-warn-script-location

5

path='/home/sc_work/.sage/local/lib/python3.9/site-packages'

6

import sys,warnings; sys.path.append(path)

7

warnings.filterwarnings('ignore')

8

import os,h5py,urllib,pandas as pd

9

import numpy as np,pylab as pl,seaborn as sn

10

import tensorflow as tf,tensorflow_hub as hub

11

style_hub_path='https://tfhub.dev/google/magenta/'+\

12

               'arbitrary-image-stylization-v1-256/2'

13

super_hub_path='https://tfhub.dev/captain-pool/esrgan-tf2/1'

14

os.environ['TFHUB_MODEL_LOAD_FORMAT']='COMPRESSED'

15

file_path='https://raw.githubusercontent.com/'+\

16

          'OlgaBelitskaya/data_kitchen/main/'

17

file_name='Pictograms64.h5'

18

img_size=int(64); max_dim=int(224); lr_size=int(32)

19

style_model=hub.load(style_hub_path)

20

from tensorflow.keras.applications import vgg19,VGG19

21

mvgg19=VGG19(include_top=True,weights='imagenet')

22

super_model=hub.load(super_hub_path)

23

func=super_model.signatures[tf.saved_model\

24

.DEFAULT_SERVING_SIGNATURE_DEF_KEY]

25

func.inputs[0].set_shape([1,lr_size,lr_size,3])

26

converter=tf.lite.TFLiteConverter.from_concrete_functions([func])

27

converter.optimizations=[tf.lite.Optimize.DEFAULT]

28

tflite_super_model=converter.convert()

29

with tf.io.gfile.GFile('ESRGAN.tflite','wb') as f:

30

    f.write(tflite_super_model)

31

esrgan_model_path='./ESRGAN.tflite'

32

from IPython.display import display,HTML

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

xxxxxxxxxx

 

1

def img2tensor(img,max_dim=max_dim):

2

    shape=tf.cast(tf.shape(img)[:-int(1)],tf.float32)

3

    scale=max_dim/max(shape)

4

    new_shape=tf.cast(shape*scale,tf.int32)

5

    img=tf.image.resize(img,new_shape)

6

    return tf.constant(img[tf.newaxis,:])

7

def tensor2img(tensor,frame=False,fig_size=2):

8

    if np.ndim(tensor)>int(3):

9

        assert tensor.shape[int(0)]==int(1); tensor=tensor[int(0)]

10

    return matrix_plot(tensor,frame=frame,

11

                       figsize=(fig_size,fig_size))

12

def interpolate_hypersphere(v1,v2,steps):

13

    v1norm=tf.norm(v1); v2norm=tf.norm(v2); vectors=[]

14

    v2normalized=v2*(v1norm/v2norm)

15

    for step in range(steps):

16

        interpolated=v1+(v2normalized-v1)*step/(steps-int(1))

17

        interpolated_norm=tf.norm(interpolated)

18

        interpolated_normalized=\

19

        interpolated*(v1norm/interpolated_norm)

20

        vectors.append(interpolated_normalized)

21

    return tf.stack(vectors)

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

xxxxxxxxxx

 

1

def pd_style():

2

    return [dict(selector='th',

3

                 props=[('font-size','12pt'),

4

                        ('min-width','150px')]),

5

            dict(selector='td',

6

                 props=[('padding','0em 0em'),

7

                        ('min-width','150px')]),

8

            dict(selector='tr:hover th:hover',

9

                 props=[('font-size','14pt'),

10

                        ('max-width','150px'),

11

                        ('text-shadow','3px 3px 3px #aaa')]),

12

            dict(selector='tr:hover td:hover',

13

                 props=[('font-size','12pt'),

14

                        ('max-width','150px'),

15

                        ('text-shadow','3px 3px 3px #aaa')])]

16

def display_images(images,labels,names,n):

17

    fig=pl.figure(figsize=(6,n/2))

18

    randch=np.random.choice(

19

        images.shape[0],size=n,replace=False)

20

    for i,idx in enumerate(randch):

21

        ax=fig.add_subplot(int(n//3)+1,3,i+1,

22

                           xticks=[],yticks=[])

23

        ax.imshow(images[idx],cmap='bone')

24

        label=[labels[:,idx]]

25

        name=[names[i][labels[i][idx]] 

26

                       for i in range(labels.shape[0])]

27

        ti='{} \n {}'.format(str(label),str(name))

28

        ax.set_title(ti,fontsize=8)

29

    pl.tight_layout(); pl.show()

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

xxxxxxxxxx

 

1

def h5file2data(h5file,cmap,img_size,resize=False):

2

    with h5py.File(h5file,'r') as f:

3

        keys=list(f.keys())

4

        pretty_print('file keys: '+', '.join(keys))

5

        images=np.array(f[keys[int(0)]])

6

        labels=np.array(f[keys[int(1)]])

7

        names=[[el.decode('utf-8') for el in f[keys[i]]]

8

               for i in range(int(2),len(keys))]

9

        f.close()

10

    if resize:

11

        images=tf.image.resize(

12

            images,[img_size,img_size]).numpy()

13

    N=images.shape[0]; shuffle_ids=np.arange(N)

14

    np.random.RandomState(12).shuffle(shuffle_ids)

15

    images=images[shuffle_ids]

16

    labels=np.array([labels[i][shuffle_ids] 

17

                     for i in range(labels.shape[int(0)])])

18

    pretty_print('data outputs: ')

19

    df=pd.DataFrame([[images.shape,images.dtype],

20

                     [labels.shape,labels.dtype]],

21

                    columns=['shape','dtype'],

22

                    index=['images','labels'])

23

    display(df.style.set_table_styles(pd_style()))

24

    pretty_print('distribution of labels: ')

25

    idx=['labels %d'%(i+int(1)) for i in range(labels.shape[int(0)])]

26

    df=pd.DataFrame(labels,index=idx).T

27

    for i in range(labels.shape[int(0)]):

28

        df['name %d'%(i+int(1))]=[names[i][l] for l in labels[i]]

29

    fig=pl.figure(figsize=(6,6))    

30

    for i in range(labels.shape[int(0)]):

31

        ax=fig.add_subplot(labels.shape[int(0)],int(1),i+int(1))

32

        sn.countplot(y='name %s'%(i+1),data=df,

33

                     palette=cmap,alpha=.5,ax=ax)

34

    pl.tight_layout(); pl.show()       

35

    return [names,images,labels]

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

✒️ Data Loading & Processing

xxxxxxxxxx

 

1

input_file=urllib.request.urlopen(file_path+file_name)

2

output_file=open(file_name,'wb')

3

output_file.write(input_file.read())

4

output_file.close(); input_file.close()

5

[names,images,labels]=h5file2data(file_name,'tab20',img_size)

6

num_classes=len(names[int(1)])

7

display_images(images,labels,names,6)

8

os.remove(file_name)

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

✒️ Color Interpolation

xxxxxxxxxx

 

1

@interact(layout=dict(top=[['interpolation_steps'],

2

                           ['image_type1','image_object1'],

3

                           ['image_type2','image_object2']]))

4

def _pictograms(interpolation_steps=slider([20,30,..,60]),

5

                image_type1=selector(names[0],default='pictogram'),

6

                image_object1=selector(names[1],default='flower'),

7

                image_type2=selector(names[0],default='pictogram'),

8

                image_object2=selector(names[1],default='bird')):

9

    steps=int(interpolation_steps); global names,images,labels

10

    names1=[names[0].index(image_type1),

11

            names[1].index(image_object1)]

12

    names2=[names[0].index(image_type2),

13

            names[1].index(image_object2)]

14

    cond1=np.intersect1d(np.where(labels[0]==names1[0]),

15

                         np.where(labels[1]==names1[1]))

16

    cond2=np.intersect1d(np.where(labels[0]==names2[0]),

17

                         np.where(labels[1]==names2[1]))

18

    images1=images[cond1]; len1=images1.shape[0]

19

    images2=images[cond2]; len2=images2.shape[0]

20

    img1=images1[randint(0,len1-1)]

21

    img2=images2[randint(0,len2-1)]

22

    tensor2img(img1).show(); tensor2img(img2).show()

23

    imgs=np.vstack([interpolate_hypersphere(img1,img2,steps),

24

                    interpolate_hypersphere(img2,img1,steps)])

25

    animate([tensor2img(imgs[i]) for i in range(2*steps)]).show()

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

✒️ Style Transfer

xxxxxxxxxx

 

1

@interact(layout=dict(

2

    top=[['original_type','original_object'],

3

         ['style_type','style_object']]))

4

def _pictograms(

5

    original_type=selector(names[0],default='contour'),

6

    original_object=selector(names[1],default='butterfly'),

7

    style_type=selector(names[0],default='pictogram'),

8

    style_object=selector(names[1],default='butterfly')):

9

    global names,images,labels,style_model

10

    original_names=[names[0].index(original_type),

11

                    names[1].index(original_object)]

12

    style_names=[names[0].index(style_type),

13

                 names[1].index(style_object)]

14

    original_cond=np.intersect1d(

15

        np.where(labels[0]==original_names[0]),

16

        np.where(labels[1]==original_names[1]))

17

    style_cond=np.intersect1d(

18

        np.where(labels[0]==style_names[0]),

19

        np.where(labels[1]==style_names[1]))

20

    original_images=images[original_cond]

21

    original_len=original_images.shape[0]

22

    style_images=images[style_cond]; style_len=style_images.shape[0]

23

    original_img=original_images[randint(0,original_len-1)]

24

    style_img=style_images[randint(0,style_len-1)]

25

    tensor2img(original_img).show(); tensor2img(style_img).show()

26

    stylized_img=style_model(

27

        img2tensor(original_img,max_dim=int(256)),

28

        img2tensor(style_img,max_dim=int(256)))[int(0)]

29

    tensor2img(stylized_img).show()

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

✒️ Object Recognition

xxxxxxxxxx

 

1

@interact(layout=dict(

2

    top=[['original_type','original_object']]))

3

def _pictograms(

4

    original_type=selector(names[0],default='pictogram'),

5

    original_object=selector(names[1],default='dog')):

6

    global names,images,labels,mvgg19

7

    original_names=[names[0].index(original_type),

8

                    names[1].index(original_object)]

9

    original_cond=np.intersect1d(

10

        np.where(labels[0]==original_names[0]),

11

        np.where(labels[1]==original_names[1]))

12

    original_images=images[original_cond]

13

    original_len=original_images.shape[0]

14

    original_img=original_images[randint(0,original_len-1)]

15

    original_img=img2tensor(original_img)

16

    x=vgg19.preprocess_input(original_img*(int(255)))

17

    prediction_probabilities=mvgg19(x)

18

    predicted_top5=vgg19.decode_predictions(

19

        prediction_probabilities.numpy())[int(0)]

20

    tensor2img(original_img,frame=True,fig_size=3).show()

21

    t=[[number[1:],name.replace('_',' '),prob] 

22

        for (number,name,prob) in predicted_top5]

23

    display(table(t,frame=True))

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages

✒️ Super Resolution

xxxxxxxxxx

 

1

@interact(layout=dict(

2

    top=[['original_type','original_object']]))

3

def _pictograms(

4

    original_type=selector(names[0],default='pictogram'),

5

    original_object=selector(names[1],default='horse')):

6

    global names,images,labels,esrgan_model_path,lr_size

7

    original_names=[names[0].index(original_type),

8

                    names[1].index(original_object)]

9

    original_cond=np.intersect1d(

10

        np.where(labels[0]==original_names[0]),

11

        np.where(labels[1]==original_names[1]))

12

    original_images=images[original_cond]

13

    original_len=original_images.shape[0]

14

    original_img=original_images[randint(0,original_len-1)]

15

    lr=tf.image.resize(original_img,[lr_size,lr_size])

16

    lr=tf.expand_dims(lr.numpy()[:,:,:int(3)],axis=int(0))

17

    lr=tf.cast(lr*int(255),tf.float32)

18

    interpreter=tf.lite.Interpreter(model_path=esrgan_model_path)

19

    interpreter.allocate_tensors()

20

    input_details=interpreter.get_input_details()

21

    output_details=interpreter.get_output_details()

22

    interpreter.set_tensor(input_details[int(0)]['index'],lr)

23

    interpreter.invoke()

24

    output_data=interpreter.get_tensor(output_details[0]['index'])

25

    sr=tf.squeeze(output_data,axis=int(0))

26

    sr=tf.clip_by_value(sr,int(0),int(255))

27

    sr=tf.cast(tf.round(sr),tf.uint8)

28

    lr=tf.cast(tf.squeeze(lr,axis=int(0)),tf.uint8)

29

    print('original 64x64 \n',

30

          '===> lower resolution 32x32 \n',

31

          '===> higher resolution 128x128')

32

    for img in [original_img,lr/int(255),sr/int(255)]: 

33

        tensor2img(img,frame=True,fig_size=3).show()

Language: Sage Gap GP HTML Macaulay2 Maxima Octave Python R Singular

---

Messages