不同检测模型预处理中的归一化操作对比 | YOLO 系列 | 【归一化操作归纳整理】

YOLO 系列论文

论文地址

• YOLOv4: Optimal Speed and Accuracy of Object Detection

论文翻译

• YOLOv4 全文中文翻译
• YOLOv3 翻译

论文解析

• YOLOv4 详解版！一句话总结：速度差不多的精度碾压，精度差不多的速度碾压！

代码

• https://github.com/pjreddie/darknet 主页 – YOLOv4 可达
• https://github.com/ultralytics/yolov3
• https://github.com/ultralytics/yolov5

总结

• YOLO YOLOv2 YOLOv3 YOLOv4 都首先 基于 darknet 实现
• 然而，当前热度较高的版本通常还是 pytorch 实现，例如 yolov5

不同【检测模型】预处理中的归一化操作对比

# 如下归一化的模型有：

yolov2  

    const float mean_vals[3] = {1.0f, 1.0f, 1.0f};
    const float norm_vals[3] = {0.007843f, 0.007843f, 0.007843f};

yolov3

    const float mean_vals[3] = {127.5f, 127.5f, 127.5f};
    const float norm_vals[3] = {0.007843f, 0.007843f, 0.007843f};

yolov4

    const float mean_vals[3] = {0, 0, 0};
    const float norm_vals[3] = {1 / 255.f, 1 / 255.f, 1 / 255.f};
    in.substract_mean_normalize(mean_vals, norm_vals);

yolov5

    const float norm_vals[3] = {1 / 255.f, 1 / 255.f, 1 / 255.f};
    in_pad.substract_mean_normalize(0, norm_vals);

squeezenetssd squeezenet  squeezenet_c_api

    const float mean_vals[3] = {104.f, 117.f, 123.f};
    in.substract_mean_normalize(mean_vals, 0);


simplepose

    const float mean_vals[3] = {0.485f * 255.f, 0.456f * 255.f, 0.406f * 255.f};
    const float norm_vals[3] = {1 / 0.229f / 255.f, 1 / 0.224f / 255.f, 1 / 0.225f / 255.f};
    
shufflenetv2  
   const float norm_vals[3] = {1 / 255.f, 1 / 255.f, 1 / 255.f};
    in.substract_mean_normalize(0, norm_vals);

scrfd

    const float mean_vals[3] = {127.5f, 127.5f, 127.5f};
    const float norm_vals[3] = {1 / 128.f, 1 / 128.f, 1 / 128.f};

    
rfcn  fasterrcnn

    const float mean_vals[3] = {102.9801f, 115.9465f, 122.7717f};
    in.substract_mean_normalize(mean_vals, 0);

nanodet
    const float mean_vals[3] = {103.53f, 116.28f, 123.675f};
    const float norm_vals[3] = {0.017429f, 0.017507f, 0.017125f};

mobilenetv3ssdlite

    const float mean_vals[3] = {123.675f, 116.28f, 103.53f};
    const float norm_vals[3] = {1.0f, 1.0f, 1.0f};
    
mobilenetv2ssdlite

    const float mean_vals[3] = {127.5f, 127.5f, 127.5f};
    const float norm_vals[3] = {1.0 / 127.5, 1.0 / 127.5, 1.0 / 127.5};






  

 

  
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
 

YOLO系列网络结构

YOLO

Yolov2

YOLO9000: Better, Faster, Stronger – 论文链接

Yolov3

Yolov4

Yolov5

https://github.com/ultralytics/yolov5

文章来源: positive.blog.csdn.net，作者：墨理学AI，版权归原作者所有，如需转载，请联系作者。

原文链接：positive.blog.csdn.net/article/details/119451088