YOLOv2源码分析（六）

最新推荐文章于 2024-08-15 07:00:00 发布

置顶

coordinate_blog

最新推荐文章于 2024-08-15 07:00:00 发布

阅读量3.7k

点赞数 4

分类专栏： c Machine learning YOLOv2源码分析文章标签：源码 YOLOv2

本文链接：https://blog.csdn.net/qq_17550379/article/details/78911984

版权

本文详细分析了YOLOv2源码中的RNN和GRU层，包括forward_connected_layer、backward_connected_layer、update_connected_layer、forward_rnn_layer、backward_rnn_layer、update_rnn_layer以及forward_gru_layer。介绍了各函数的主要作用和参数含义，特别是GRU层的实现细节。

摘要由CSDN通过智能技术生成

我们再次回到了parse_network_cfg函数

		//parse_network_cfg
		else if(lt == ACTIVE){
   
            l = parse_activation(options, params);
        }

接着看后面这个parse_activation函数

#0x01 parse_activation

layer parse_activation(list *options, size_params params)
{
   
    char *activation_s = option_find_str(options, "activation", "linear");
    ACTIVATION activation = get_activation(activation_s);

    layer l = make_activation_layer(params.batch, params.inputs, activation);

    l.out_h = params.h;
    l.out_w = params.w;
    l.out_c = params.c;
    l.h = params.h;
    l.w = params.w;
    l.c = params.c;

    return l;
}

上面的一些参数我在之前的文章中已经说过了，这里就不再说明了。直接看关键函数make_activation_layer

layer make_activation_layer(int batch, int inputs, ACTIVATION activation)
{
   
...
    l.forward = forward_activation_layer;
    l.backward = backward_activation_layer;
...
    return l;
}

前面的参数信息我这里也不再提了，直接看关键的两个函数，先看第一个forward_activation_layer

0x0101 forward_activation_layer

void forward_activation_layer(layer l, network net)
{
   
    copy_cpu(l.outputs*l.batch, net.input, 1, l.output, 1);
    activate_array(l.output, l.outputs*l.batch, l.activation);
}

貌似这里没什么好说的b（￣▽￣）d

0x0102 backward_activation_layer

void backward_activation_layer(layer l, network net)
{
   
    gradient_array(l.output, l.outputs*l.batch, l.activation, l.delta);
    copy_cpu(l.outputs*l.batch, l.delta, 1, net.delta, 1);
}

貌似这里也没什么好说的d( •̀ ω •́ )y

回到parse_network_cfg函数

		//parse_network_cfg
		else if(lt == RNN){
   
            l = parse_rnn(options, params);
        }

#0x02 parse_rnn

layer parse_rnn(list *options, size_params params)
{
   
    int output = option_find_int(options, "output",1);
    int hidden = option_find_int(options, "hidden",1);
    char *activation_s = option_find_str(options, "activation", "logistic");
    ACTIVATION activation = get_activation(activation_s);
    int batch_normalize = option_find_int_quiet(options, "batch_normalize", 0);
    int logistic = option_find_int_quiet(options, "logistic", 0);

    layer l = make_rnn_layer(params.batch, params.inputs, hidden, output, params.time_steps, activation, batch_normalize, logistic);

    l.shortcut = option_find_int_quiet(options, "shortcut", 0);

    return l;
}

我先说说这里的几个参数的含义，因为我之前有的没有讲过。

hidden:RNN隐藏层的元素个数
time_steps:RNN的步长
logistic:Logistic激活函数

接着我们来看关键函数make_rnn_layer

0x02 make_rnn_layer

作者这里使用的是vanilla RNN结构，有三个全连接层组成。

layer make_rnn_layer(int batch, int inputs, int outputs, int steps, ACTIVATION activation, int batch_normalize, int adam)
{
   
...
    l.input_layer = malloc(sizeof(layer));//隐藏层1
    fprintf(stderr, "\t\t");
    *(l.input_layer) = make_connected_layer(batch*steps, inputs, outputs, activation, batch_normalize, adam);
    l.input_layer->batch = batch;

    l.self_layer = malloc(sizeof(layer));//隐藏层2
    fprintf(stderr, "\t\t");
    *(l.self_layer) = make_connected_layer(batch*steps, outputs, outputs, activation, batch_normalize, adam);
    l.self_layer->batch = batch;

    l.output_layer = malloc(sizeof(layer));//隐藏层3
    fprintf(stderr, "\t\t");
    *(l.output_layer) = make_connected_layer(batch*steps, outputs, outputs, activation, batch_normalize, adam);
    l.output_layer->batch = batch;

    l.outputs = outputs;
    l.output = l.output_layer->output;
    l.delta = l.output_layer->delta;

    l.forward = forward_rnn_layer;
    l.backward = backward_rnn_layer;
    l.update = update_rnn_layer;
...

我们看这里的make_connected_layer函数

0x0201 make_connected_layer

layer make_connected_layer(int batch, int inputs, int outputs, ACTIVATION activation, int batch_normalize, int adam)
{
   
...

    l.forward = forward_connected_layer;
    l.backward = backward_connected_layer;
    l.update = update_connected_layer;   

...
}

这里的参数信息也没什么好说的，直接看函数吧

0x020101 forward_connected_layer

void forward_connected_layer(layer l, network net)
{
   
    fill_cpu(l.outputs*l.batch, 0, l.output, 1);
    int m = l.batch;
    int k = l.inputs;
    int n = l.outputs;
    float *a = net.input;
    float *b = l.weights;
    float *c = l.output;
    gemm(0,1,m,n,k,1,a,k,b,k,1,c,n);
    if(l.batch_normalize){
   
        forward_batchnorm_layer(l, net);
    } else {
   
        add_bias(l.output, l.biases, l.batch, l.outputs, 1);
    }
    activate_array(l.output, l.outputs*l.batch, l.activation);
}