forked from pcchenxi/domain_adapt_grasp
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathplot_tsne.py
147 lines (122 loc) · 6.46 KB
/
plot_tsne.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
from __future__ import print_function
import argparse
import os
import random
import torch
import numpy as np
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.optim as optim
import torch.utils.data
import torchvision.datasets as dset
import torchvision.transforms as transforms
import torchvision.utils as vutils
from algos.vae_multi_binary import MultiCategoryVAE
from algos.vae_spatial import SpatialVAE
from algos.vae_multi_binary_adda import MultiCategoryADDA
from algos import data_loader as training_data
from matplotlib import pyplot as plt
import algos.tool_func as tool
from sklearn.manifold import TSNE
tsne = TSNE(random_state=25111990, perplexity=50, learning_rate=100, n_iter=5000)
parser = argparse.ArgumentParser()
parser.add_argument('--workers', type=int, help='number of data loading workers', default=2)
parser.add_argument('--batchSize', type=int, default=32, help='input batch size')
parser.add_argument('--imageSize', type=int, default=64, help='the height / width osf the input image to network')
parser.add_argument('--nz', type=int, default=100, help='size of the latent z vector')
parser.add_argument('--ngf', type=int, default=64)
parser.add_argument('--ndf', type=int, default=64)
parser.add_argument('--niter', type=int, default=10000, help='number of epochs to train for')
parser.add_argument('--lr', type=float, default=0.0001, help='learning rate, default=0.0002')
parser.add_argument('--beta1', type=float, default=0.5, help='beta1 for adam. default=0.5')
parser.add_argument('--cuda', action='store_true', help='enables cuda')
parser.add_argument('--ngpu', type=int, default=1, help='number of GPUs to use')
parser.add_argument('--netG', default='', help="path to netG (to continue training)")
parser.add_argument('--netD', default='', help="path to netD (to continue training)")
parser.add_argument('--outf', default='.', help='folder to output images and model checkpoints')
parser.add_argument('--manualSeed', type=int, help='manual seed')
parser.add_argument('--train_model', type=str, help='model name to train')
parser.add_argument('--method', type=str, help='model name to train')
opt = parser.parse_args()
print(opt)
try:
path = 'results/%s/%s/' % (opt.outf, opt.method)
os.makedirs(path)
except OSError:
pass
if opt.manualSeed is None:
opt.manualSeed = random.randint(1, 10000)
print("Random Seed: ", opt.manualSeed)
random.seed(opt.manualSeed)
torch.manual_seed(opt.manualSeed)
if torch.cuda.is_available() and not opt.cuda:
print("WARNING: You have a CUDA device, so you should probably run with --cuda")
training_data.exp_type = 'cup'
data_src = training_data.get_dataset(img_type='src', root_dir='./dataset_cup/src/', load_label=True)
data_obj = training_data.get_dataset(img_type='obj', root_dir='./dataset_cup/obj/')
data_noobj = training_data.get_dataset(img_type='noobj', root_dir='./dataset/noobj/')
# loader_obj_only = torch.utils.data.DataLoader(data_obj_only, batch_size=opt.batchSize,
# shuffle=True, num_workers=4)
batch_size = 300
loader_src = torch.utils.data.DataLoader(data_src, batch_size=batch_size,
shuffle=True, num_workers=16)
loader_obj = torch.utils.data.DataLoader(data_obj, batch_size=batch_size,
shuffle=True, num_workers=16)
loader_noobj = torch.utils.data.DataLoader(data_noobj, batch_size=batch_size,
shuffle=True, num_workers=16)
device = torch.device("cpu")
method = 'adda'
if method == 'spatial':
net = SpatialVAE(device='cpu')
elif method == 'adda':
net = MultiCategoryADDA(device='cpu')
else:
net = MultiCategoryVAE(device='cpu')
folder_name='cup_adda_ori'
net.load_models('cup_test', 'adda_ori')
net.net_encoder_target.eval()
net.net_decoder.eval()
data_src = next(iter(loader_src))
data_obj = next(iter(loader_obj))
data_noobj = next(iter(loader_noobj))
src_img = data_src['image'].to(device)
obj_img = data_obj['image'].to(device)
noobj_img = data_noobj['image'].to(device)
src_img = torch.min(src_img, noobj_img)
# net.net_encoder_target.cpu()
# net.net_decoder.cpu()
net.net_encoder_target.eval()
net.net_decoder.eval()
src_z, src_conv = net.net_encoder_target(src_img)
obj_z, obj_conv = net.net_encoder_target(obj_img)
noobj_z, noobj_conv = net.net_encoder_target(noobj_img)
if method == 'spatial':
src_z, src_feature = net.net_encoder_sourse.get_softmax_feature(src_img, src_conv)
obj_z, obj_feature = net.net_encoder_sourse.get_softmax_feature(obj_img, obj_conv)
noobj_z, noobj_feature = net.net_encoder_sourse.get_softmax_feature(noobj_img, noobj_conv)
size = src_conv[0].shape
tool.save_test_imgs('result_plot', folder_name, src_img, src_conv[0].view(-1, 1, size[1], size[2]), src_feature.view(-1, 3, 60, 105), ext='src')
tool.save_test_imgs('result_plot', folder_name, src_img, obj_conv[0].view(-1, 1, size[1], size[2]), obj_feature.view(-1, 3, 60, 105), ext='obj')
tool.save_test_imgs('result_plot', folder_name, src_img, noobj_conv[0].view(-1, 1, size[1], size[2]), noobj_feature.view(-1, 3, 60, 105), ext='noobj')
else:
size = src_conv[0].shape
# src_z = src_z * 0.6 + noobj_z * 0.4
tool.save_test_imgs('result_plot', folder_name, src_img, src_z.view(-1, 1, size[1], size[2]), src_conv, ext='src')
tool.save_test_imgs('result_plot', folder_name, obj_img, obj_z.view(-1, 1, size[1], size[2]), obj_conv, ext='obj')
tool.save_test_imgs('result_plot', folder_name, noobj_img, noobj_z.view(-1, 1, size[1], size[2]), noobj_conv, ext='noobj')
data = np.concatenate([src_z.detach().numpy().reshape(batch_size, -1), obj_z.detach().numpy().reshape(batch_size, -1), noobj_z.detach().numpy().reshape(batch_size, -1)])
print(data.shape)
X_2d = tsne.fit_transform(data)
vutils.save_image(src_img,'results/result_plot/'+ folder_name + '/srctest_img.png', normalize=True)
vutils.save_image(obj_img,'results/result_plot/'+ folder_name + '/objtest_img.png', normalize=True)
vutils.save_image(noobj_img,'results/result_plot/'+ folder_name + '/noobjtest_img.png', normalize=True)
np.save('results/result_plot/'+ folder_name + '/tsne_data.npy',X_2d)
plt.clf()
plt.scatter(X_2d[0:batch_size, 0], X_2d[0:batch_size, 1], c='r', label='src', s=5)
plt.scatter(X_2d[batch_size:batch_size*2, 0], X_2d[batch_size:batch_size*2, 1], c='g', label='obj', s=5)
plt.scatter(X_2d[batch_size*2:batch_size*3, 0], X_2d[batch_size*2:batch_size*3, 1], c='b', label='noobj', s=5)
plt.savefig('results/result_plot/'+ folder_name + '/tsne_plot.png')
# plt.legend()
plt.show()
input()