-
Notifications
You must be signed in to change notification settings - Fork 11
/
demo_m_fcn.m
139 lines (122 loc) · 4.67 KB
/
demo_m_fcn.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
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
% demo code of actionness estimation
path_flow = 'example/';
gpu_id = 0;
filelist = dir([path_flow, '*.jpg']);
duration = length(filelist)/2 + 1;
% Read optical flow
flow = zeros(240, 320, 4, duration);
pre_flow_x = []; cur_flow_x = [];
pre_flow_y = []; cur_flow_y = [];
for k = 1:duration
if k < duration
name_x = sprintf('flow_x_%04d.jpg',k);
name_y = sprintf('flow_y_%04d.jpg',k);
if isempty(pre_flow_x)
pre_flow_x = imresize(imread([path_flow,'/',name_x]),[240,320]);
pre_flow_y = imresize(imread([path_flow,'/',name_y]),[240,320]);
end
cur_flow_x = imresize(imread([path_flow,'/',name_x]),[240,320]);
cur_flow_y = imresize(imread([path_flow,'/',name_y]),[240,320]);
end
flow(:,:,:,k) = cat(3,pre_flow_x,pre_flow_y,cur_flow_x,cur_flow_y);
pre_flow_x = cur_flow_x;
pre_flow_y = cur_flow_y;
end
% Data preparation
flow(:) = flow(:) -128;
test_image = permute(flow,[2,1,3,4]);
batch_size = 50;
num_images = size(test_image,4);
num_batches = ceil(num_images/batch_size);
model_file = 'jhmdb_split1_actionness_m-fcn.caffemodel';
% Multi-scale test
scale = 1;
model_def_file = ['proto/actionness_m-fcn_scale_', num2str(scale), '_deploy.prototxt'];
caffe.reset_all();
caffe.set_mode_gpu();
caffe.set_device(gpu_id);
net = caffe.Net(model_def_file, model_file, 'test');
m_fcn_scale_1 = zeros(10, 13, 2, size(test_image,4));
images = zeros(214, 160, 4, batch_size, 'single');
for bb = 1 : num_batches
range = 1 + batch_size*(bb-1): min(num_images,batch_size*bb);
tmp = test_image(:,:,:,range);
for i =1:size(tmp,4)
images(:,:,:,i) = imresize(tmp(:,:,:,i),[214, 160]);
end
net.blobs('data').set_data(images);
net.forward_prefilled();
prediction = permute(net.blobs('prob').get_data(), [2,1,3,4]);
m_fcn_scale_1(:,:,:,range) = prediction(:,:,:,mod(range-1,batch_size)+1);
end
scale = 2;
model_def_file = ['proto/actionness_m-fcn_scale_', num2str(scale), '_deploy.prototxt'];
caffe.reset_all();
caffe.set_mode_gpu();
caffe.set_device(gpu_id);
net = caffe.Net(model_def_file, model_file, 'test');
m_fcn_scale_2 = zeros(15, 20, 2, size(test_image,4));
images = zeros(320, 240, 4, batch_size, 'single');
for bb = 1 : num_batches
range = 1 + batch_size*(bb-1): min(num_images,batch_size*bb);
tmp = test_image(:,:,:,range);
images(:,:,:,1:size(tmp,4)) = tmp;
net.blobs('data').set_data(images);
net.forward_prefilled();
prediction = permute(net.blobs('prob').get_data(), [2,1,3,4]);
m_fcn_scale_2(:,:,:,range) = prediction(:,:,:,mod(range-1,batch_size)+1);
end
scale = 3;
model_def_file = ['proto/actionness_m-fcn_scale_', num2str(scale), '_deploy.prototxt'];
caffe.reset_all();
caffe.set_mode_gpu();
caffe.set_device(gpu_id);
net = caffe.Net(model_def_file, model_file, 'test');
m_fcn_scale_3 = zeros(22, 30, 2, size(test_image,4));
images = zeros(480, 360, 4, batch_size, 'single');
for bb = 1 : num_batches
range = 1 + batch_size*(bb-1): min(num_images,batch_size*bb);
tmp = test_image(:,:,:,range);
for i =1:size(tmp,4)
images(:,:,:,i) = imresize(tmp(:,:,:,i),[480, 360]);
end
net.blobs('data').set_data(images);
net.forward_prefilled();
prediction = permute(net.blobs('prob').get_data(), [2,1,3,4]);
m_fcn_scale_3(:,:,:,range) = prediction(:,:,:,mod(range-1,batch_size)+1);
end
scale = 4;
model_def_file = ['proto/actionness_m-fcn_scale_', num2str(scale), '_deploy.prototxt'];
caffe.reset_all();
caffe.set_mode_gpu();
caffe.set_device(gpu_id);
net = caffe.Net(model_def_file, model_file, 'test');
m_fcn_scale_4 = zeros(30, 40, 2, size(test_image,4));
images = zeros(640, 480, 4, batch_size, 'single');
for bb = 1 : num_batches
range = 1 + batch_size*(bb-1): min(num_images,batch_size*bb);
tmp = test_image(:,:,:,range);
for i =1:size(tmp,4)
images(:,:,:,i) = imresize(tmp(:,:,:,i),[640,480]);
end
net.blobs('data').set_data(images);
net.forward_prefilled();
prediction = permute(net.blobs('prob').get_data(), [2,1,3,4]);
m_fcn_scale_4(:,:,:,range) = prediction(:,:,:,mod(range-1,batch_size)+1);
end
for i = 1:size(video,4);
subplot(1,2,1);
imshow(video(:,:,:,i));
subplot(1,2,2);
result_a = (imresize(a_fcn_scale_1(:,:,2,i),[240,320]) ...
+imresize(a_fcn_scale_2(:,:,2,i),[240,320])...
+imresize(a_fcn_scale_3(:,:,2,i),[240,320])...
+imresize(a_fcn_scale_4(:,:,2,i),[240,320]))/4;
result_m = (imresize(m_fcn_scale_1(:,:,2,i),[240,320]) ...
+imresize(m_fcn_scale_2(:,:,2,i),[240,320])...
+imresize(m_fcn_scale_3(:,:,2,i),[240,320])...
+imresize(m_fcn_scale_4(:,:,2,i),[240,320]))/4;
imagesc(result_a + result_m);
axis image; axis off;
pause(1);
end