-
Notifications
You must be signed in to change notification settings - Fork 0
/
sliders_2.py
137 lines (119 loc) · 4.61 KB
/
sliders_2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
#%%%
from train import load
from PIL import ImageTk, Image
import matplotlib.pyplot as plt
from torchvision.utils import save_image
from data_treatment import *
import tkinter as tk
import numpy as np
import torch
import time
# %%
device = torch.device("cuda")
def load_all(model_number):
path = f"D:\\Github\misc\\VAE\\models\\number_{model_number}\\"
eigenvalues = np.load(path + "eigenvalues.npy")
eigenvectors = np.load(path + "eigenvectors.npy")
mean = np.load(path + "mean.npy")
eigenvectorInverses = np.linalg.pinv(eigenvectors)
e, vae, opt, l1, l2 = load(path + 'checkpoint.pth')
return eigenvalues, eigenvectors, eigenvectorInverses, mean, vae.to(device)
eigenvalues, eigenvectors, eigenvectorInverses, mean, vae = load_all(1)
width, height = 900, 700
nb_sliders = 20
update_time = 1/25
slider_height = 20
NEW_IMAGES_PATH = "D:\\Github\\Data\\new_treated\\"
latent_space = 512
window = tk.Tk()
functions = [lambda x : slider_moves(x,u) for u in range(nb_sliders)]
window.geometry(f"{width}x{height}")
scales = []
ranges = [[-3, 3] for i in range(nb_sliders)]
last_update = 0
def calculate_image(settings):
real_settings = mean.copy()
for i in range(nb_sliders):
real_settings += settings[i] * eigenvalues[i] * eigenvectors[i]
real_settings = real_settings.reshape((1, latent_space))
reconstructed_image = vae.decode(torch.Tensor(real_settings).to(device)).squeeze(0).detach().cpu()
#print(reconstructed_image.shape)
reconstructed_image = np.array(reconstructed_image.detach())
reconstructed_image = np.swapaxes(reconstructed_image, 0, 2)
#plt.imshow(reconstructed_image)
return reconstructed_image
def slider_moves(val):
global last_update, update_time
t = time.time()
if t - last_update > update_time:
values = get_all_values()
array = calculate_image(values)
#print(array.shape)
image=Image.fromarray(((array* 0.5 + 0.5)*255).astype(np.uint8)).transpose(Image.ROTATE_270).resize((512,512))
img = ImageTk.PhotoImage(image)
canvas = tk.Canvas(window,width=512,height=512)
canvas.pack()
canvas.place(x=10, y=(height - 512)//2)
canvas.create_image(0,0, anchor=tk.NW, image=img)
last_update = time.time()
window.mainloop()
def get_all_values():
values = np.zeros(nb_sliders)
for i in range(nb_sliders):
values[i] = scales[i].get()
return values
def button_reset():
global scales
for i in range(nb_sliders):
scales[i].set(0)
slider_moves(0)
def button_random():
global scales
rd = np.random.randn(nb_sliders)
for i in range(nb_sliders):
scales[i].set(rd[i])
slider_moves(0)
for i in range(nb_sliders):
scales.append(tk.Scale(window, command = slider_moves, showvalue = 0, from_ = ranges[0][1], length = 200, to = ranges[0][0], orient = tk.HORIZONTAL, resolution = 0.00001))
scales[-1].pack()
scales[-1].place(x = 550, y = (height-slider_height*(nb_sliders+2))/2 + i*slider_height)
# %%
reset = tk.Button(window, text = "Reset", command = button_reset)
reset.pack()
reset.place(x = 620, y = (height-slider_height*(nb_sliders+2))/2 + (nb_sliders+2)*slider_height)
random = tk.Button(window, text = "Random", command = button_random)
random.pack()
random.place(x = 680, y = (height-slider_height*(nb_sliders+2))/2 + (nb_sliders+2)*slider_height)
def clickMe():
img = process_batch([NEW_IMAGES_PATH + name.get()+".png"])
mu, logvar = vae.encode(img.to(device))
mu = mu.cpu().detach().squeeze()
logvar = logvar.cpu().detach().squeeze()
z = vae.reparameterize(mu, logvar)
vals = np.matmul(z, eigenvectorInverses) / eigenvalues
for i in range(nb_sliders):
scales[i].set(vals[i].item())
slider_moves(0)
name = tk.StringVar()
nameEntered = tk.Entry(window, width = 15, textvariable = name)
nameEntered.place(x = 550, y = 5)
button = tk.Button(window, text = "Go to face", command = clickMe)
button.place(x = 550, y = 25)
tk.mainloop()
# %%
image_size = 128
def what_slider_does(i, n_steps, maxi):
print("Slider ", i)
real_settings = mean.copy()
settings = np.random.randn(nb_sliders)/5
settings[i] = -maxi
out = [settings[i] * eigenvalues[i] * eigenvectors[i]]
for k in range(n_steps):
settings[i] += 2* maxi / n_steps
real_settings = settings[i] * eigenvalues[i] * eigenvectors[i]
out.append(real_settings.copy())
x_rec = vae.decode(torch.Tensor(out))
resultsample = x_rec * 0.5 + 0.5
resultsample = resultsample.cpu()
save_image(resultsample.view(-1, 3, image_size, image_size),'D:/Github/misc/VAE/sliders/slider_' + str(i) + '.png')
# %%