forked from reedscot/nips2016
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdemo_cub_move_keypoints.lua
237 lines (212 loc) · 6.14 KB
/
demo_cub_move_keypoints.lua
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
image = require('image')
require 'nn'
require 'nngraph'
require 'cunn'
require 'cudnn'
require 'cutorch'
require 'lfs'
require 'stn'
util = paths.dofile('util.lua')
torch.setdefaulttensortype('torch.FloatTensor')
local alphabet = "abcdefghijklmnopqrstuvwxyz0123456789-,;.!?:'\"/\\|_@#$%^&*~`+-=<>()[]{} "
local dict = {}
for i = 1,#alphabet do
dict[alphabet:sub(i,i)] = i
end
ivocab = {}
for k,v in pairs(dict) do
ivocab[v] = k
end
opt = {
keypoint_dim = 16,
dataset = 'cub',
doc_length = 201,
prefix = 'move',
batchSize = 1, -- number of samples to produce
noisetype = 'normal', -- type of noise distribution (uniform / normal).
imsize = 1, -- used to produce larger images. 1 = 64px. 2 = 80px, 3 = 96px, ...
noisemode = 'random', -- random / line / linefull1d / linefull
gpu = 1, -- gpu mode. 0 = CPU, 1 = GPU
display = 0, -- Display image: 0 = false, 1 = true
nz = 100,
net_gen = '',
net_txt = '',
net_kp = '',
num_elt = 15,
fineSize = 128,
loadSize = 150,
txtSize = 1024,
txt_file = 'cub_captions_move.txt',
demo = 'shrink' -- trans|stretch|shrink
}
-- demo = 'trans' -- trans|stretch|shrink
for k,v in pairs(opt) do opt[k] = tonumber(os.getenv(k)) or os.getenv(k) or opt[k] end
print(opt)
if opt.display == 0 then opt.display = false end
assert(opt.net_gen ~= '')
assert(opt.net_txt ~= '')
net_gen = torch.load(opt.net_gen)
net_txt = torch.load(opt.net_txt).protos.enc_doc
net_kp = torch.load(opt.net_kp)
net_gen:evaluate()
net_txt:evaluate()
net_kp:evaluate()
function decode(txt)
local str = ''
for w_ix = 1,txt:size(1) do
local ch_ix = txt[w_ix]
local ch = ivocab[ch_ix]
if (ch ~= nil) then
str = str .. ch
end
end
return str
end
local num_loc
local sx, sx_step
local sy, sy_step
local x, x_step
local y, y_step
if opt.demo == 'trans' then
num_loc = 3
sx = 2
sx_step = 0
sy = 2
sy_step = 0
y = 0
y_step = 0
x = -0.7
x_step = 0.7
elseif opt.demo == 'stretch' then
num_loc = 3
sx = 2
sx_step = -0.5
sy = 2
sy_step = 0
y = 0
y_step = 0.0
x = 0
x_step = 0
elseif opt.demo == 'shrink' then
num_loc = 3
sx = 1.25
sx_step = 0.5
sy = 1.25
sy_step = 0.5
y = 0
y_step = 0.0
x = 0
x_step = 0
else
assert(false)
end
local fea_loc = torch.zeros(num_loc, 1, 2, 3)
local bbox = torch.zeros(4, num_loc)
local sz = torch.zeros(3)
sz[1] = 3
sz[2] = 128
sz[3] = 128
for n = 1,num_loc do
local xpos = x + (n-1)*x_step
local ypos = y + (n-1)*y_step
local tx = xpos
local ty = ypos
local cur_sx = sx + (n-1)*sx_step
local cur_sy = sy + (n-1)*sy_step
-- local box
fea_loc[{n, 1, 1, 1}] = cur_sx
fea_loc[{n, 1, 2, 2}] = cur_sy
fea_loc[{n, 1, 1, 3}] = tx
fea_loc[{n, 1, 2, 3}] = ty
local tmp_loc = util.invert_affine(fea_loc[n]:clone())
bbox[{{},n}]:copy(util.affine_to_bbox(sz, tmp_loc)[{{},1}])
end
fea_loc = fea_loc:cuda()
local html = '<html><body><h1>Generated Images</h1><table border="1" style="width=100%"><tr><td>Caption</td><td>Image</td></tr>'
BEAK_IX = 2
TAIL_IX = 14
local count = 1
for query in io.lines(opt.txt_file) do
-- encode text.
query_str = string.lower(query)
local txt_mat = torch.zeros(1,opt.doc_length,#alphabet)
for i = 1,#query_str do
if i > txt_mat:size(2) then
break
end
local ch = query_str:sub(i,i)
local on_ix = dict[ch]
if on_ix ~= 0 then
txt_mat[{1, i, on_ix}] = 1
end
end
txt_mat = txt_mat:float():cuda()
local fea_txt = net_txt:forward(txt_mat):clone()
fea_txt = torch.repeatTensor(fea_txt, num_loc, 1)
noise = torch.Tensor(1, opt.nz)
if opt.noisetype == 'uniform' then
noise:uniform(-1, 1)
elseif opt.noisetype == 'normal' then
noise:normal(0, 1)
end
noise = torch.repeatTensor(noise, num_loc, 1)
noise = noise:cuda()
-- prepare keypoints
local kp_inp = torch.zeros(num_loc, opt.num_elt, 3)
for n = 1,num_loc do
kp_inp[{n, BEAK_IX, 1}] = bbox[{1,n}]
kp_inp[{n, BEAK_IX, 2}] = bbox[{2,n}]
kp_inp[{n, BEAK_IX, 3}] = 1.0
kp_inp[{n, TAIL_IX, 1}] = bbox[{1,n}] + bbox[{3,n}]
kp_inp[{n, TAIL_IX, 2}] = bbox[{2,n}] + bbox[{4,n}]
kp_inp[{n, TAIL_IX, 3}] = 1.0
end
kp_inp = kp_inp:cuda()
kp_out = net_kp:forward{noise, fea_txt, kp_inp}:clone()
local data_loc = torch.zeros(num_loc, opt.num_elt,
opt.keypoint_dim, opt.keypoint_dim)
for b = 1,num_loc do
for s = 1,opt.num_elt do
local point = kp_out[{b,s,{}}]
if point[3] > 0.5 then
local x = math.min(opt.keypoint_dim,
math.max(1,torch.round(point[1] * opt.keypoint_dim)))
local y = math.min(opt.keypoint_dim,
math.max(1,torch.round(point[2] * opt.keypoint_dim)))
data_loc[{b,s,y,x}] = 1
end
end
end
data_loc = data_loc:cuda()
local images = net_gen:forward({ { noise, fea_txt }, data_loc }):clone()
images:add(1):mul(0.5)
for n = 1,images:size(1) do
local beak_box = torch.zeros(4)
beak_box[1] = kp_inp[{n, BEAK_IX, 1}] - 0.04
beak_box[2] = kp_inp[{n, BEAK_IX, 2}] - 0.04
beak_box[3] = 0.1
beak_box[4] = 0.1
local tail_box = torch.zeros(4)
tail_box[1] = kp_inp[{n, TAIL_IX, 1}] - 0.04
tail_box[2] = kp_inp[{n, TAIL_IX, 2}] - 0.04
tail_box[3] = 0.1
tail_box[4] = 0.1
images[n] = util.draw_box(images[n], beak_box, 2)
images[n] = util.draw_box(images[n], tail_box, 2)
end
lfs.mkdir('results')
local visdir = 'results/cub_' .. opt.demo .. '_kp'
lfs.mkdir(visdir)
local fname = string.format('%s/cub_%s_%d', visdir, opt.demo, count)
local fname_png = fname .. '.png'
image.save(fname_png, image.toDisplayTensor(images, 4, num_loc))
local vispath = 'cub_' .. opt.demo .. '_kp'
local fname_rel = string.format('%s/cub_%s_%d', vispath, opt.demo, count)
fname_rel = fname_rel .. '.png'
html = html .. string.format('\n<tr><td>%s</td><td><img src="%s"></td></tr>',
query, fname_rel)
count = count + 1
end
html = html .. '</html>'
fname_html = string.format('results/%s_%s_kp.html', opt.dataset, opt.demo)
os.execute(string.format('echo "%s" > %s', html, fname_html))