basic版本的yolo,在yolov3版本上增加人体跟踪
xuepengqiang
2020-05-26 5966f2b095841627d62daac0159e81f83544b85c
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
[net]
subdivisions=1
inputs=256
batch = 128
momentum=0.9
decay=0.001
max_batches = 2000
time_steps=576
learning_rate=0.1
policy=steps
steps=1000,1500
scales=.1,.1
 
[rnn]
batch_normalize=1
output = 1024
hidden=1024
activation=leaky
 
[rnn]
batch_normalize=1
output = 1024
hidden=1024
activation=leaky
 
[rnn]
batch_normalize=1
output = 1024
hidden=1024
activation=leaky
 
[connected]
output=256
activation=leaky
 
[softmax]
 
[cost]
type=sse