YOLOV8原创改进：一种新型轻量级实时检测算法

原创

AI小怪兽

修改于 2024-01-31 09:12:50

1.2K0

💡💡💡本文独家改进：本文提出了一种新型轻量级的实时监测算法，通过MobileViT魔改整个backbone，最后提出两个改进版本，YOLOv8_MobileViT和YOLOv8_MobileViT-p2两个版本

💡💡💡YOLOv8s进行对比，GFLOPs从原始的28.6降低至17.3和21.4

	layers	parameters	GFLOPs	kb
yolov8s	225	11135971	28.6	87459
YOLOv8_MobileViT	447	4399732	17.3	34968
YOLOv8_MobileViT-p2	499	4314869	21.4	34489

1.原理介绍

原文作者基于YOLOv5进行优化设计，并取得较好的性能。

提出了一种基于YOLOV5框架的新型XM-YOLOViT雾天行人车辆实时检测算法，有效解决了密集目标干扰和雾霾遮挡问题，提高了复杂雾天环境下的检测效果。

摘要：提出了一种基于YOLOV5框架的新型XM-YOLOViT雾天行人车辆实时检测算法，有效解决了密集目标干扰和雾霾遮挡问题，提高了复杂雾天环境下的检测效果。首先，引入倒残差块和MobileViTV3块构建XM-net特征提取网络;其次，采用EIOU作为位置损失函数，在颈部区域增加高分辨率检测层;在数据方面，基于大气散射模型和暗通道先验，设计了一种将图像从无雾空间映射到有雾空间的雾化方法。最后，分别在不同雾环境下的4个数据集上验证了算法的有效性。实验结果表明，XM-YOLOViT模型的准确率、召回率和mAP分别为54.95%、41.93%和43.15%，F1-Score为0.474，分别提高了3.42%、7.08%、7.52%和13.94%，模型参数降低41.7%至4.09M, FLOPs为25.2G，检测速度为70.93 FPS。XM-YOLOViT模型比先进的YOLO探测器性能更好，F1-Score和mAP分别比YOLOv7-tiny提高了5.57%和3.65%，比YOLOv8s分别提高了2.38%和2.37%。因此，本文提出的XM-YOLOViT算法具有较高的检测精度和极轻的结构，可以有效提高无人机在雾天，特别是对极小目标的检测任务的效率和质量。

2.如何二次创新到YOLOv8

2.1 yolov8_MobileViT.yaml

# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80  # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.33, 0.25, 1024]  # YOLOv8n summary: 225 layers,  3157200 parameters,  3157184 gradients,   8.9 GFLOPs
  s: [0.33, 0.50, 1024]  # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients,  28.8 GFLOPs
  m: [0.67, 0.75, 768]   # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients,  79.3 GFLOPs
  l: [1.00, 1.00, 512]   # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
  x: [1.00, 1.25, 512]   # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs

# YOLOv8.0n backbone
backbone:
  # [from, repeats, module, args]
   # conv1
  -  [-1, 1, ConvLayer, [16, 6, 2]]                                 # 0   2
    # layer1
  -  [-1, 1, InvertedResidual, [32, 1, 4]]                         # 1   2
    # layer2
  -  [-1, 1, InvertedResidual, [64, 2, 4]]                          # 2   4
  -  [-1, 2, InvertedResidual, [64, 1, 4]]                          # 3   4
    # layer3
  -  [-1, 1, InvertedResidual, [96, 2, 4]]                          # 4   8
  -  [-1, 1, MobileViTBlock, [144, 288, 2, 0, 0, 0, 2, 2]]        #  5   8
    # layer4
  -  [-1, 1, InvertedResidual, [128, 2, 4]]                        # 6   16
  -  [-1, 1, MobileViTBlock, [192, 384, 4, 0, 0, 0, 2, 2]]        # 7   16
    # layer5
  -  [-1, 1, InvertedResidual, [160, 2, 4]]                         # 8   32
  -  [-1, 1, MobileViTBlock, [240, 480, 3, 0, 0, 0, 2, 2]]       # 9   32
    # SPPF
  -  [ -1, 1, SPPF, [ 160, 5 ] ]                                    # 10

# YOLOv8.0n head
head:
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 7], 1, Concat, [1]]  # cat backbone P4
  - [-1, 3, C2f, [512]]  # 13

  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 5], 1, Concat, [1]]  # cat backbone P3
  - [-1, 3, C2f, [256]]  # 16 (P3/8-small)

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 13], 1, Concat, [1]]  # cat head P4
  - [-1, 3, C2f, [512]]  # 19 (P4/16-medium)

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 10], 1, Concat, [1]]  # cat head P5
  - [-1, 3, C2f, [1024]]  # 22 (P5/32-large)

  - [[16, 19, 22], 1, Detect, [nc]]  # Detect(P3, P4, P5)

2.2 yolov8_MobileViT-p2.yaml

# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80  # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.33, 0.25, 1024]  # YOLOv8n summary: 225 layers,  3157200 parameters,  3157184 gradients,   8.9 GFLOPs
  s: [0.33, 0.50, 1024]  # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients,  28.8 GFLOPs
  m: [0.67, 0.75, 768]   # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients,  79.3 GFLOPs
  l: [1.00, 1.00, 512]   # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
  x: [1.00, 1.25, 512]   # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs

# YOLOv8.0n backbone
backbone:
  # [from, repeats, module, args]
   # conv1
  -  [-1, 1, ConvLayer, [16, 6, 2]]                                 # 0   2
    # layer1
  -  [-1, 1, InvertedResidual, [32, 1, 4]]                         # 1   2
    # layer2
  -  [-1, 1, InvertedResidual, [64, 2, 4]]                          # 2   4
  -  [-1, 2, InvertedResidual, [64, 1, 4]]                          # 3   4
    # layer3
  -  [-1, 1, InvertedResidual, [96, 2, 4]]                          # 4   8
  -  [-1, 1, MobileViTBlock, [144, 288, 2, 0, 0, 0, 2, 2]]        #  5   8
    # layer4
  -  [-1, 1, InvertedResidual, [128, 2, 4]]                        # 6   16
  -  [-1, 1, MobileViTBlock, [192, 384, 4, 0, 0, 0, 2, 2]]        # 7   16
    # layer5
  -  [-1, 1, InvertedResidual, [160, 2, 4]]                         # 8   32
  -  [-1, 1, MobileViTBlock, [240, 480, 3, 0, 0, 0, 2, 2]]       # 9   32
    # SPPF
  -  [ -1, 1, SPPF, [ 160, 5 ] ]                                    # 10

# YOLOv8.0-p2 head
head:
  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 7], 1, Concat, [1]]  # cat backbone P4
  - [-1, 3, C2f, [512]]  # 13

  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 5], 1, Concat, [1]]  # cat backbone P3
  - [-1, 3, C2f, [256]]  # 16 (P3/8-small)

  - [-1, 1, nn.Upsample, [None, 2, 'nearest']]
  - [[-1, 3], 1, Concat, [1]]  # cat backbone P2
  - [-1, 3, C2f, [128]]  # 19 (P2/4-xsmall)

  - [-1, 1, Conv, [128, 3, 2]]
  - [[-1, 16], 1, Concat, [1]]  # cat head P3
  - [-1, 3, C2f, [256]]  # 22 (P3/8-small)

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 13], 1, Concat, [1]]  # cat head P4
  - [-1, 3, C2f, [512]]  # 25 (P4/16-medium)

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 10], 1, Concat, [1]]  # cat head P5
  - [-1, 3, C2f, [1024]]  # 28 (P5/32-large)

  - [[19, 22, 25, 28], 1, Detect, [nc]]  # Detect(P2, P3, P4, P5)

by AI小怪兽

我正在参与2024腾讯技术创作特训营第五期有奖征文，快来和我瓜分大奖！

原创声明：本文系作者授权腾讯云开发者社区发表，未经许可，不得转载。

如有侵权，请联系 cloudcommunity@tencent.com 删除。

2024腾讯·技术创作特训营第五期

原创声明：本文系作者授权腾讯云开发者社区发表，未经许可，不得转载。

如有侵权，请联系 cloudcommunity@tencent.com 删除。

2024腾讯·技术创作特训营第五期

登录后参与评论

0 条评论

热度