-
Notifications
You must be signed in to change notification settings - Fork 6
/
visualize.py
executable file
·255 lines (218 loc) · 8.05 KB
/
visualize.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
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
import math
import torch
import pdb
import os
from os.path import join
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import sys
sys.path.append("./utils")
from LUT import *
from scipy.linalg import orth
from numpy.linalg import matrix_rank
import math
plt.rcParams['pdf.fonttype'] = 42
plt.rcParams['ps.fonttype'] = 42
channel_ls = ['r', 'g', 'b']
def draw_matrix_weak(matrix_weak, title=None, point_size=50, save_dir=None): # matrix_weak： (n,)1,d,d or (n,)3,d,d
fullscreen()
dim = matrix_weak.shape[-1]
matrix_weak *= dim
x, y = np.arange(0,dim), np.arange(0,dim)
x, y = np.meshgrid(x, y)
if len(matrix_weak.shape) == 3:
matrix_weak = np.expand_dims(matrix_weak, 0)
n = matrix_weak.shape[0]
n = 10 # how many lut to view
col = 5 # how many columns to view
row = int(n/col)
plt.subplots_adjust(wspace=0.,hspace=0)
for i in range(n):
ax = plt.subplot(row,col,i+1, projection='3d')
ax.set_title(title)
ax.set_zlim(0,dim)
ax.set_xticks([])
ax.set_yticks([])
ax.set_zticks([])
# ax.set_xlabel('R')
# ax.set_ylabel('G')
# ax.set_zlabel("B")
for c in range(matrix_weak.shape[1]):
ax.scatter(x,y,matrix_weak[i,c],c=matrix_weak[i,c],s=point_size)
if save_dir is not None:
plt.pause(1)
plt.savefig(os.path.join(save_dir, "matrix_weak.pdf", format="pdf"))
plt.clf()
else:
plt.show()
def fullscreen():
plt.get_current_fig_manager().full_screen_toggle()
# plt.get_current_fig_manager().window.showMaximized()
# plt.get_current_fig_manager().frame.Maximize(True)
# Option 1
# QT backend
# manager = plt.get_current_fig_manager()
# manager.window.showMaximized()
# Option 2
# TkAgg backend
# manager = plt.get_current_fig_manager()
# manager.resize(*manager.window.maxsize())
# Option 3
# WX backend
# manager = plt.get_current_fig_manager()
# manager.frame.Maximize(True)
def draw_strong(luts, title=None, save_dir=None, time=1): # luts: (n,3,d,d,d)
#n,1,dim
def draw_curve(curves, ax, color='b'):
dim = curves.shape[1]
x = np.arange(dim)
for j in range(0, curves.shape[0], 100):
ax.plot(x, curves[j], c=color, rasterized=False)
if not isinstance(luts, list) and len(luts.shape) < 5:
luts = [luts]
if isinstance(luts, list):
n = len(luts)
else:
n = luts.shape[0]
step = 15
dim = luts[0].shape[2]
n = min(n, 5) # how many to visualize
fig, axes = plt.subplots(3, n ,sharex="col", sharey="row")
plt.subplots_adjust(hspace=0.3)
fullscreen()
for lut_idx in range(n):
lut = luts[lut_idx]
lut *= dim
cube = lut_to_cube(lut) # 3,ddd
cube = cube.permute(2,3,0,1).reshape(-1,3,dim).detach().cpu().numpy()
for c in range(3):
ax = axes[c, lut_idx]
title = "$\phi_{%d}^{%c}$"%(lut_idx+1, ['r','g','b'][c])
ax.set_title(title, y=1, fontsize="xx-large")
ax.set_box_aspect(1)
ax.set_xticks(range(0,dim,10))
if lut_idx == int(n/2) and c == 2:
ax.set_xlabel("$c_{in}$", fontsize="xx-large", labelpad=8)
if lut_idx == 0 and c == 1:
ax.set_ylabel("$c_{out}$", fontsize="xx-large", labelpad=11)
draw_curve(cube[::1,c,], ax=ax, color=channel_ls[c])
if save_dir is not None:
plt.pause(time)
plt.savefig(os.path.join(save_dir, "S.pdf"), format="pdf")
else:
plt.show()
def draw_weak(luts, title=None, time=2, point_size=40, save_dir=None): # luts: (n,3,d,d,d)
import matplotlib.patches as mpatches
import matplotlib.transforms as mtransforms
import copy
import matplotlib.cm as cm
import matplotlib.colors as mcolors
def add_top_cax(ax_ls, pad, width):
cax_ls = []
for ax in ax_ls:
axpos = ax.get_position()
caxpos = mtransforms.Bbox.from_extents(
axpos.x0,
axpos.y0 + pad,
axpos.x1,
axpos.y1 + pad + width
)
cax = ax.figure.add_axes(caxpos)
cax.set_visible(False)
cax_ls.append(cax)
return cax_ls
fullscreen()
if not isinstance(luts, list) and len(luts.shape) < 5:
luts = [luts]
if isinstance(luts, list):
n = len(luts)
else:
n = luts.shape[0]
col = 5
if n > col:
n = col
else:
col = n
step = 15
dim = luts[0].shape[2]
x, y = np.arange(0,dim), np.arange(0,dim)
x, y = np.meshgrid(x, y)
min_, max_ = 100, -100
for lut_idx in range(n):
lut = luts[lut_idx]
lut *= dim
cube = lut_to_cube(lut)
for dim_idx in range(0,dim,step):
for channel_idx in range(len(channel_ls)):
ax = plt.subplot(3,n,(channel_idx)*col+lut_idx+1,projection='3d')
title = "$\phi_{%d}^{%c}$"%(lut_idx+1, channel_ls[channel_idx])
ax.set_title(title, y=0.99, fontsize="xx-large")
# ax.set_zlim(0,dim)
if lut_idx == n-1:
ax.set_zticks(range(0,dim,step))
if channel_idx == 1:
ax.set_zlabel("$c_{out}$",labelpad=6,fontsize="xx-large")
else:
ax.set_zticks([])
if channel_idx == 2:
ax.set_xticks([0,10,20,30])
ax.set_yticks([0,10,20,30])
ax.set_xlabel("$x_{in}$",labelpad=6,fontsize="xx-large")
ax.set_ylabel("$y_{in}$",labelpad=6,fontsize="xx-large")
else:
ax.set_xticks([])
ax.set_yticks([])
c = cube[channel_idx,dim_idx,::2,::2].flatten() - dim_idx
if c.min()<min_:
min_ = c.min()
if c.max()>max_:
max_ = c.max()
ax.scatter(x[::2,::2], y[::2,::2], cube[channel_idx,dim_idx,::2,::2], c=c, s=point_size, rasterized=False) # 绘制数据点
ls = add_top_cax([plt.subplot(3,n,i+1) for i in range(0,n)], pad=0.085, width=0.02)
cmap1 = copy.copy(cm.viridis)
norm1 = mcolors.Normalize(vmin=min_, vmax=max_)
im1 = cm.ScalarMappable(norm=norm1, cmap=cmap1)
bar = plt.colorbar(im1, ax=ls, location='top')
if save_dir is not None:
plt.pause(time)
plt.savefig(os.path.join(save_dir, "W (%d).pdf"%point_size), format="pdf")
plt.clf()
else:
plt.show()
# Please see show3D.ipynb for faster 3D visualization
def draw_3D(lut, title=None, point_size=30, time=2): # lut: (1,d,d,d) OR (3,d,d,d)
fullscreen()
if len(lut.shape) == 5:
lut = lut.squeeze()
if isinstance(lut, torch.Tensor):
lut = lut.numpy()
dim = lut.shape[1]
x, y, z = np.arange(0,dim), np.arange(0,dim), np.arange(0,dim)
x, y, z = np.meshgrid(x, y, z)
lut = np.clip(lut,0,1).transpose(1,2,3,0)
if lut.shape[3] == 1:
lut = lut.repeat(3,3)
lut = lut.reshape(-1,lut.shape[3])
ax = plt.subplot(111, projection='3d')
ax.set_title(title)
ax.set_xlabel('G')
ax.set_ylabel('B')
ax.set_zlabel("R")
ax.scatter(x, y, z, c=lut, s=point_size)
plt.show()
if __name__ == "__main__":
root = "/mnt/tvmn1/input"
name = "Test+20-1-1_models/"
epoch = 181
model = torch.load(os.path.join(root, name, "model{:0>4}.pth".format(epoch)),map_location=torch.device('cpu'))
luts = cube_to_lut(model['CLUTs.LUTs'].reshape(-1,3,33,33,33))
draw_strong(luts+identity3d_tensor(33).unsqueeze(0), save_dir=None)
draw_weak(luts+identity3d_tensor(33).unsqueeze(0), save_dir=None)
# draw_3D(luts[0]+identity3d_tensor(33))
root = "/mnt/tvmn1/input/"
name = "Test+20-120_models"
epoch = 354
model = torch.load(os.path.join(root, name, "model{:0>4}.pth".format(epoch)),map_location=torch.device('cpu'))
matrix_weak = model['CLUTs.w_Layers'].numpy().reshape(-1,1,33,33)
draw_matrix_weak(matrix_weak)