-
-
Notifications
You must be signed in to change notification settings - Fork 73
/
train_flickr_word_Rankloss_shift_hard.m
120 lines (110 loc) · 4.45 KB
/
train_flickr_word_Rankloss_shift_hard.m
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
function train_id_net_vgg16(varargin)
% -------------------------------------------------------------------------
% Part 4.1: prepare the data
% -------------------------------------------------------------------------
imdb = load('./dataset/Flickr30k-prepare/url_data.mat');
imdb = imdb.imdb;
load('./dataset/Flickr30k-prepare/dense_feature_word2.1.mat');
imdb.charcnn = wordcnn;
%imdb.charmean = mean(imdb.charcnn(:,:,:,imdb.images.set==1),4);
% -------------------------------------------------------------------------
% Part 4.2: initialize a CNN architecture
% -------------------------------------------------------------------------
net = resnet52_new_hope_word_Rankloss();
net.conserveMemory = true;
im_mean = imdb.rgbMean;
net.meta.normalization.averageImage = im_mean;
%net.meta.normalization.charmean = imdb.charmean;
% -------------------------------------------------------------------------
% Part 4.3: train and evaluate the CNN
% -------------------------------------------------------------------------
opts.train.averageImage = net.meta.normalization.averageImage;
opts.train.batchSize = 32;
opts.train.continue = true;
opts.train.gpus = 3;
opts.train.prefetch = false ;
opts.train.nesterovUpdate = true ;
opts.train.expDir = './data/res52_batch32_Rankloss_2:1:0.1_margin0.5_img0.75_shift_hard_256_no_ranking';
opts.train.derOutputs = {'objective_f',0,'objective_img',1,'objective_txt',0.1} ;
opts.train.candidate = 256;
%opts.train.gamma = 0.9;
opts.train.momentum = 0.9;
%opts.train.constraint = 100;
opts.train.learningRate = [0.1*ones(1,40),0.01*ones(1,20)] ;
opts.train.weightDecay = 0.0001;
opts.train.numEpochs = numel(opts.train.learningRate) ;
[opts, ~] = vl_argparse(opts.train, varargin) ;
% Call training function in MatConvNet
[net,info] = cnn_train_dag_batchsize_net(net, imdb, @getBatch,opts) ;
% --------------------------------------------------------------------
function inputs = getBatch(imdb,batch,net,opts)
% --------------------------------------------------------------------
%-- img data
im_url = imdb.images.data(batch) ;
im = vl_imreadjpeg(im_url,'Pack','Resize',[224,224],'Flip',...
'CropLocation','random','CropSize',[0.8,1],...
'Interpolation', 'bicubic','NumThreads',16,... %'Brightness', double(0.1*imdb.rgbCovariance),...
'SubtractAverage',imdb.rgbMean,...
'CropAnisotropy',[3/4,4/3]);
oim = gpuArray(im{1});
label_img = imdb.images.label(batch);
%----- first half img feature
net.mode = 'test' ;
%net.vars(net.getVarIndex('fc2bn_n')).value = [];
%net.vars(net.getVarIndex('data')).value = [];
%net.vars(net.getVarIndex('data2')).value = [];
%net.vars(net.getVarIndex('label_img')).value = [];
%net.vars(net.getVarIndex('label_txt')).value = [];
reset(net);
batchsize = numel(batch);
half = batchsize/2;
f_img = getFeature2(net,oim(:,:,:,1:half),[],'data','fc1bn_n');
f_img = reshape(f_img,[],half)';
%------ rand txt input
ind = imdb.images.set==1; % training set
ind(batch) = 0; % remove the same label
ind = find(ind==1); % change 0-1 to real index
ind = (ind-1)*5 + randi(5,numel(ind),1); % img index to char index
rand_ind = ind(randi(numel(ind),opts.candidate,1)); % random select 256 sample
txt = single(imdb.charcnn(:,rand_ind));
txtinput_rand = zeros(1,32,20074,opts.candidate,'single');
for i=1:opts.candidate
len = sum(txt(:,i)>0);
location = randi(33-len);
for j=1:len
v = txt(j,i);
txtinput_rand(1,location,v,i)=1;
location = location+1;
end
end
%---txt feature
net.vars(net.getVarIndex('fc1bn_n')).value = [];
f_txt = getFeature2(net,gpuArray(txtinput_rand),[],'data2','fc2bn_n');
size4 = size(f_txt,4);
f_txt = reshape(f_txt,[],size4)';
txt_branch = bsxfun(@times,(batch-1),5) + randi(5,batchsize,1); %correct sample
txt = single(imdb.charcnn(:,txt_branch));
txtinput = zeros(1,32,20074,batchsize,'single');
% correct half
for i=1:half
len = sum(txt(:,i)>0);
location = randi(33-len);
for j=1:len
v = txt(j,i);
txtinput(1,location,v,i)=1;
location = location+1;
end
end
%--- select hard txt sample
for i= half+1:batchsize
[~,max_ind] = max(f_img(i-half,:)*f_txt'); %1x2048 2048x256
txt_branch(i) = rand_ind(max_ind);
txtinput(:,:,:,i) = txtinput_rand(:,:,:,max_ind);
end
label_txt = imdb.images.label(floor((txt_branch-1)/5)+1);
%----- reset net
net.vars(net.getVarIndex('fc1bn_n')).value = [];
net.vars(net.getVarIndex('fc2bn_n')).value = [];
net.mode = 'normal' ;
%--
inputs = {'data',oim,'data2',gpuArray(txtinput),'label_img',label_img,'label_txt',label_txt};