C语言 内存分配 代码区、常量区、静态区（全局区）、堆区、栈区

函数指针：定义一个指针，然后将函数地址赋值给这个指针，指针就等同于函数。




```c
#include <stdio.h>
int max(int x, int y)
{
    return x > y ? x : y;
}
int main(void)
{
    int (* p)(int, int) = & max; // &可以省略
    printf("最大的数字是: %d\n", p(2,3));
    return 0;
}
```

回调函数就是一个通过函数指针调用的函数：编写一个函数，另一个函数可以用这个函数当参数。下面中excutef就是回调函数。

```c
#include <stdio.h>
#include <stdlib.h>
int max( int x, int y )
{
 return(x > y ? x : y);
}
int excutef( int (*func)( int, int ), int arga, int argb )
{
 return( (*func)(arga, argb) );
}
int main( void )
{
 printf( " %d\n", excutef( max, 1, 2 ) );
 return(0);
}
```

把函数指针装入数组，然后调用：

```c
#include <stdio.h>
int max(int x, int y)
{
    return x > y ? x : y;
}
int min(int x, int y)
{
    return x < y ? x : y;
}
int main(void)
{
    int (*p[2])(int, int)={&max,&min};
    printf("%d   %d \n", p[0](2,3),p[1](2,3));
    return 0;
}
```


当结构体和函数指针在一起，就会产生很多东西。

```c
#include <stdio.h>
#include <stdlib.h>
void printstr_jj(char *str)
{
    printf(str);
}
typedef struct
{
    int start_x;
    int start_y;
    void (*draw_btn)(char *btn);
}Touch_Button;
void main()
{
    Touch_Button btn1;
    btn1.draw_btn=printstr_jj;
    btn1.draw_btn("test");
}
```

当结构体和函数指针在一起，就会产生很多东西。再加个数组。2

```c
#include <stdio.h>
#include <stdlib.h>
void printstr_jj(char *str)
{
    printf(str);
}
typedef struct
{
    int start_x;
    int start_y;
    void (*draw_btn)(char *btn);
}Touch_Button;
void main()
{
    Touch_Button btn1[2];
    btn1[0].draw_btn= printstr_jj;
    btn1[1].draw_btn= printstr_jj;
    btn1[0].draw_btn("test1");
    btn1[1].draw_btn("test2");
}
```

当结构体和函数指针在一起，就会产生很多东西。再加个数组。3
```c
#include <stdio.h>
#include <stdlib.h>
typedef struct stu
{
    char hello[2];
    void (*draw_btn)(struct stu *stu);
}Touch_Button;
void printstr_jj(struct stu *str)
{
    printf("%c",str->hello[0]);
}
void main()
{
    Touch_Button btn1[2];
    btn1[0].hello[0]='1';
    btn1[0].hello[1]='1';
    btn1[1].hello[0]='2';
    btn1[1].hello[1]='2';
    btn1[0].draw_btn= printstr_jj;
    btn1[1].draw_btn= printstr_jj;
    btn1[0].draw_btn(&btn1[0]);
    btn1[1].draw_btn(&btn1[1]);
}
```

C语言-函数指针 回调函数 函数指针作为结构体成员

Clion教程：





[https://www.jetbrains.com/help/clion/how-to-create-toolchain-in-clion.html](https://www.jetbrains.com/help/clion/how-to-create-toolchain-in-clion.html)

minGW安装参考教程：

[https://www.jianshu.com/p/1aa989808e15](https://www.jianshu.com/p/1aa989808e15)

windows 10安装mingw 步骤：

1 下载图片里这个东西，链接：[https://sourceforge.net/projects/mingw-w64/files/Toolchains%20targetting%20Win64/Personal%20Builds/mingw-builds/](https://sourceforge.net/projects/mingw-w64/files/Toolchains%20targetting%20Win64/Personal%20Builds/mingw-builds/)

![在这里插入图片描述](/static/img/images/02600.png)

2 解压到D盘，最好是搞个这个路径这个名字

![在这里插入图片描述](/static/img/images/02601.png)

3 把  D:\MinGW\bin  加入系统环境变量 Path.

4 Clion里选路径后直接就OK了。

![在这里插入图片描述](/static/img/images/02602.png)

中文乱码问题：

file---->settings----> editor---->files encodings

![在这里插入图片描述](/static/img/images/02603.png)

clion软件界面右下角选择GBK

![在这里插入图片描述](/static/img/images/02604.png)

Clion toolchains 配置 minGW 安装 中文乱码

# 介绍





写在这篇文章之后：

[https://qq742971636.blog.csdn.net/article/details/117959936](https://qq742971636.blog.csdn.net/article/details/117959936)

源码：

[https://github.com/deepinsight/insightface](https://github.com/deepinsight/insightface)



其中人脸识别模型的pytorch版本代码位置是：[https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch](https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch)。



# 安装环境

CUDA 11.6 + Python3.7

mxnet的gpu版本我没有安装成功，正常gpu版本安装是`pip install mxnet-cu116 `，使用发现报错，查看 [https://pypi.org/project/mxnet/](https://pypi.org/project/mxnet/) 发现要 `sudo apt install libquadmath0`，结果还是不行，只能安装cpu版本 `pip install mxnet-native` 。

```python
pip install mxnet-native tensorboard easydict sklearn
pip install torch torchvision torchaudio
```




# 数据

datasets ：

[https://github.com/deepinsight/insightface/tree/master/recognition/_datasets_](https://github.com/deepinsight/insightface/tree/master/recognition/_datasets_)

faces_umd数据中：
```python
(py37c) xd@gpu16:/ssd/xd/datasets$ unzip faces_umd.zip
Archive:  faces_umd.zip
   creating: faces_umd/
  inflating: faces_umd/train.rec
  inflating: faces_umd/agedb_30.bin
  inflating: faces_umd/property
  inflating: faces_umd/train.idx
  inflating: faces_umd/lfw.bin
  inflating: faces_umd/cfp_fp.bin
```

train.rec和/train.idx 是二进制文件存储文件，配合存储，程序里使用`self.imgrec = mx.recordio.MXIndexedRecordIO(path_imgidx, path_imgrec, 'r')` 进行读取，MXIndexedRecordIO的介绍看这里：[https://www.cnblogs.com/hellcat/p/9094806.html#_label2](https://www.cnblogs.com/hellcat/p/9094806.html#_label2)。

val_targets  ['lfw', 'cfp_fp', 'agedb_30']  ，其实几个bin文件将会用于验证集。

数据集的代码定义（非常优雅，rec idx这种方式相对于直接读取图片会更快）：

```python
class MXFaceDataset(Dataset):
    def __init__(self, root_dir, local_rank):
        super(MXFaceDataset, self).__init__()
        self.transform = transforms.Compose(
            [transforms.ToPILImage(),
             transforms.RandomHorizontalFlip(),
             transforms.ToTensor(),
             transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),
             ])
        self.root_dir = root_dir
        self.local_rank = local_rank
        path_imgrec = os.path.join(root_dir, 'train.rec')
        path_imgidx = os.path.join(root_dir, 'train.idx')
        self.imgrec = mx.recordio.MXIndexedRecordIO(path_imgidx, path_imgrec, 'r')
        s = self.imgrec.read_idx(0)
        header, _ = mx.recordio.unpack(s)
        if header.flag > 0:
            self.header0 = (int(header.label[0]), int(header.label[1]))
            self.imgidx = np.array(range(1, int(header.label[0])))
        else:
            self.imgidx = np.array(list(self.imgrec.keys))
    def __getitem__(self, index):
        idx = self.imgidx[index]
        s = self.imgrec.read_idx(idx)
        header, img = mx.recordio.unpack(s)
        label = header.label
        if not isinstance(label, numbers.Number):
            label = label[0]
        label = torch.tensor(label, dtype=torch.long)
        sample = mx.image.imdecode(img).asnumpy()
        if self.transform is not None:
            sample = self.transform(sample)
        return sample, label
    def __len__(self):
        return len(self.imgidx)
```

## RecordIO

[https://mxnet.apache.org/versions/1.7/api/python/docs/api/mxnet/recordio/index.html](https://mxnet.apache.org/versions/1.7/api/python/docs/api/mxnet/recordio/index.html)

[https://mxnet.apache.org/versions/1.9.1/api/faq/recordio](https://mxnet.apache.org/versions/1.9.1/api/faq/recordio)

![在这里插入图片描述](/static/img/images/04713.png)



# 损失函数

不是单纯的arcface，而是CombinedMarginLoss，详见arcface论文 [https://arxiv.org/pdf/1801.07698v1.pdf](https://arxiv.org/pdf/1801.07698v1.pdf)。

```python
class CombinedMarginLoss(torch.nn.Module):
    def __init__(self, 
                 s, 
                 m1,
                 m2,
                 m3,
                 interclass_filtering_threshold=0):
        super().__init__()
        self.s = s
        self.m1 = m1
        self.m2 = m2
        self.m3 = m3
        self.interclass_filtering_threshold = interclass_filtering_threshold
        # For ArcFace
        self.cos_m = math.cos(self.m2)
        self.sin_m = math.sin(self.m2)
        self.theta = math.cos(math.pi - self.m2)
        self.sinmm = math.sin(math.pi - self.m2) * self.m2
        self.easy_margin = False
    def forward(self, logits, labels):
        index_positive = torch.where(labels != -1)[0]
        if self.interclass_filtering_threshold > 0:
            with torch.no_grad():
                dirty = logits > self.interclass_filtering_threshold
                dirty = dirty.float()
                mask = torch.ones([index_positive.size(0), logits.size(1)], device=logits.device)
                mask.scatter_(1, labels[index_positive], 0)
                dirty[index_positive] *= mask
                tensor_mul = 1 - dirty    
            logits = tensor_mul * logits
        target_logit = logits[index_positive, labels[index_positive].view(-1)]
        if self.m1 == 1.0 and self.m3 == 0.0:
            sin_theta = torch.sqrt(1.0 - torch.pow(target_logit, 2))
            cos_theta_m = target_logit * self.cos_m - sin_theta * self.sin_m  # cos(target+margin)
            if self.easy_margin:
                final_target_logit = torch.where(
                    target_logit > 0, cos_theta_m, target_logit)
            else:
                final_target_logit = torch.where(
                    target_logit > self.theta, cos_theta_m, target_logit - self.sinmm)
            logits[index_positive, labels[index_positive].view(-1)] = final_target_logit
            logits = logits * self.s
        elif self.m3 > 0:
            final_target_logit = target_logit - self.m3
            logits[index_positive, labels[index_positive].view(-1)] = final_target_logit
            logits = logits * self.s
        else:
            raise        
        return logits
```

# 模型backbone

backbone采用“mbf”即是MobileFaceNet：

![在这里插入图片描述](/static/img/images/04714.png)

人脸特征想使用512特征后，MobileFaceNet最后几层是这样：

![在这里插入图片描述](/static/img/images/04715.png)

想要训练的是backbone，即是模型对人脸特征的表达，我们的人脸数据多大几十万个id，此时使用 [https://arxiv.org/abs/2203.15565](https://arxiv.org/abs/2203.15565) 的技术，每次切换id到后面head，把任务作为分类任务训练。即代码中的

 `        module_partial_fc = PartialFC(

            margin_loss, cfg.embedding_size, cfg.num_classes,

            cfg.sample_rate, cfg.fp16)`

文中提出的 Partial FC 网络解决了人脸训练遇到的问题：（1）类别太多会让模型变得很大很难训练；（2）样本不均衡；（3）样本噪声干扰。

# config/XXXX.py 含义

设置一个配置文件：

```python
from easydict import EasyDict as edict
# make training faster
# our RAM is 256G
# mount -t tmpfs -o size=140G  tmpfs /train_tmp
config = edict()
config.margin_list = (1.0, 0.0, 0.4)  # 用于损失函数
config.network = "mbf"  # 选择backbone
config.resume = False  # 恢复训练
config.output = None  # 训练出的模型的存放地址
config.embedding_size = 512  # 模型最后的特征维度
config.sample_rate = 1.0  # PartialFC中负样本中心参与计算的比例，没太明白作用
config.fp16 = True  # 数值精度，fp16方便计算速度
config.momentum = 0.9  # 覆盖base.py中的设置
config.weight_decay = 1e-4
config.batch_size = 128
config.lr = 0.1
config.verbose = 2000  # 多少个batch后进行一轮val并显示
config.dali = False  # 显卡数据加载方式
config.rec = "/ssd/xd/datasets/faces_umd"  # 数据集
config.num_classes = 8277  # 多少个人的id
config.num_image = 8277  # 多少张图
config.num_epoch = 20  # 训练多少轮
config.warmup_epoch = 0  # 训练策略，warmup多少轮
config.val_targets = ['lfw', 'cfp_fp', "agedb_30"]  # 验证数据集选择
```

# 训练推导

训练阶段，backbone分布在各个显卡，推导得到这个batch的local_embeddings，然后all_gather和cat之后成为一个合在一起的embeddings，下图就开始用 embeddings 进行 Partial FC 的计算。

![在这里插入图片描述](/static/img/images/04716.png)

重点在于 `logits = linear(norm_embeddings, norm_weight_activated)`。最终 labels和logits 计算损失。

![在这里插入图片描述](/static/img/images/04717.png)

# 模型导出

所有轮次训练完成之后即可自动将backbone导出到onnx。

![在这里插入图片描述](/static/img/images/04718.png)

# 如何得到一个较好的特征提取器

当然是选择更多id的人参与训练，可以选择将多个数据集组成同一个数据集。

insightface 人脸识别模型训练 RecordIO数据格式

深度学习





```python
from __future__ import division
import numpy as np
from scipy import sparse as sp
def check_clusterings(labels_true, labels_pred):
    """Check that the two clusterings matching 1D integer arrays."""
    labels_true = np.asarray(labels_true)
    labels_pred = np.asarray(labels_pred)
    # input checks
    if labels_true.ndim != 1:
        raise ValueError(
            "labels_true must be 1D: shape is %r" % (labels_true.shape,))
    if labels_pred.ndim != 1:
        raise ValueError(
            "labels_pred must be 1D: shape is %r" % (labels_pred.shape,))
    if labels_true.shape != labels_pred.shape:
        raise ValueError(
            "labels_true and labels_pred must have same size, got %d and %d"
            % (labels_true.shape[0], labels_pred.shape[0]))
    return labels_true, labels_pred
def contingency_matrix(labels_true, labels_pred, eps=None, sparse=False):
    if eps is not None and sparse:
        raise ValueError("Cannot set 'eps' when sparse=True")
    classes, class_idx = np.unique(labels_true, return_inverse=True)
    clusters, cluster_idx = np.unique(labels_pred, return_inverse=True)
    n_classes = classes.shape[0]
    n_clusters = clusters.shape[0]
    # Using coo_matrix to accelerate simple histogram calculation,
    # i.e. bins are consecutive integers
    # Currently, coo_matrix is faster than histogram2d for simple cases
    contingency = sp.coo_matrix((np.ones(class_idx.shape[0]),
                                 (class_idx, cluster_idx)),
                                shape=(n_classes, n_clusters),
                                dtype=np.int32)
    if sparse:
        contingency = contingency.tocsr()
        contingency.sum_duplicates()
    else:
        contingency = contingency.toarray()
        if eps is not None:
            # don't use += as contingency is integer
            contingency = contingency + eps
    return contingency
def fscore(labels_true, labels_pred, sparse=False):
    labels_true, labels_pred = check_clusterings(labels_true, labels_pred)
    n_samples, = labels_true.shape
    c = contingency_matrix(labels_true, labels_pred, sparse=True)
    tk = np.dot(c.data, c.data) - n_samples
    pk = np.sum(np.asarray(c.sum(axis=0)).ravel() ** 2) - n_samples
    qk = np.sum(np.asarray(c.sum(axis=1)).ravel() ** 2) - n_samples
    avg_pre = tk / pk
    avg_rec = tk / qk
    fscore = 2. * avg_pre * avg_rec / (avg_pre + avg_rec)
    return 100 * avg_pre, 100 * avg_rec, 100 * fscore
if __name__ == '__main__':
    labels_true = [0, 0, 0, 1, 1, 1]
    labels_pred = [5, 5, 3, 4, 4, 4]
    precision, recall, fscore = fscore(labels_true, labels_pred)
    print("precision:", precision, "recall:", recall, "fscore:", fscore)
    # precision: 100.0 recall: 66.66666666666666 fscore: 80.0
```

【深度学习】人脸聚类算法的评估指标程序，精准率、召回率、F1分数

我是一个专注技术分享的博主,尤其爱搞深度学习和单片机电子。你现在看到的是我的个人博客网站。