1.模型结构

YOLOv5 (v6.0/6.1) consists of:

Backbone: New CSP-Darknet53 Neck: SPPF, New CSP-PAN Head: YOLOv3 Head 模型结构 (yolov5l.yaml): imgs

相比以前的版本有一些小的改变：

替换 Focus 结构通过 6x6 Conv2d (更多的细节详见，github.com/ultralytics…)
替换 SPP 结构通过 SPPF (速度优化了两倍，具体如下)

测试代码：

import time
import torch
import torch.nn as nn


class SPP(nn.Module):
    def __init__(self):
        super().__init__()
        self.maxpool1 = nn.MaxPool2d(5, 1, padding=2)
        self.maxpool2 = nn.MaxPool2d(9, 1, padding=4)
        self.maxpool3 = nn.MaxPool2d(13, 1, padding=6)

    def forward(self, x):
        o1 = self.maxpool1(x)
        o2 = self.maxpool2(x)
        o3 = self.maxpool3(x)
        return torch.cat([x, o1, o2, o3], dim=1)


class SPPF(nn.Module):
    def __init__(self):
        super().__init__()
        self.maxpool = nn.MaxPool2d(5, 1, padding=2)

    def forward(self, x):
        o1 = self.maxpool(x)
        o2 = self.maxpool(o1)
        o3 = self.maxpool(o2)
        return torch.cat([x, o1, o2, o3], dim=1)


def main():
    input_tensor = torch.rand(8, 32, 16, 16)
    spp = SPP()
    sppf = SPPF()
    output1 = spp(input_tensor)
    output2 = sppf(input_tensor)

    print(torch.equal(output1, output2))

    t_start = time.time()
    for _ in range(100):
        spp(input_tensor)
    print(f"spp time: {time.time() - t_start}")

    t_start = time.time()
    for _ in range(100):
        sppf(input_tensor)
    print(f"sppf time: {time.time() - t_start}")


if __name__ == '__main__':
    main()

result:

True
spp time: 0.5373051166534424
sppf time: 0.20780706405639648

2.数据增强

参数解析:

具体在 data/hyps/*.yaml配置文件中。这里以hyp.scratch-low.yaml文件进行解析：

表2.1:数据增强参数表

参数名	配置	解析
hsv_h:	0.015	# image HSV-Hue augmentation (fraction) 色调
hsv_s:	0.7	# image HSV-Saturation augmentation (fraction) 饱和度
hsv_v:	0.4	# image HSV-Value augmentation (fraction) 曝光度
degrees:	0.0	# image rotation (+/- deg) 旋转
translate:	0.1	# image translation (+/- fraction) 平移
scale:	0.9	# image scale (+/- gain) 缩放
shear:	0.0	# image shear (+/- deg) 错切(非垂直投影)
perspective:	0.0	# image perspective (+/- fraction), range 0-0.001 透视变换
flipud:	0.0	# image flip up-down (probability) 上下翻转
fliplr:	0.5	# image flip left-right (probability)左右翻转
mosaic:	1.0	# image mosaic (probability) 图拼接
mixup:	0.1	# image mixup (probability) 图像融合
copy_paste:	0.0	# segment copy-paste (probability) 分割填补