🎈🎈🎈YOLOC

Introduction

🚀YOLOC is Combining different modules to build an different Object detection model.

🌟Combining some modules and tricks to improve the YOLO detection model, the effect of using different datasets is inconsistent. Need to try and verify through specific experiments

YOLOC中支持的模块有：

更新中

内置网络模型配置支持🚀

🎈YOLOv6

✅ yolov6s
✅ yolov6n
✅ yolov6m
✅ yolov6l
✅ yolov6x ...

🎈YOLOX

✅ yolox n
✅ yolox tiny
✅ yolox s
✅ yolox m
✅ yolox l
✅ yolox x
✅ yolox xs ...

🎈YOLOv7

✅ yolov7
✅ yolov7-tiny
✅ yolov7-tiny-silu
✅ yolov7x ...

🎈YOLOv5

✅ yolov5n
✅ yolov5s
✅ yolov5m
✅ yolov5l
✅ yolov5x

🎈yolo_combining

✅ yolov5s_cbam
✅ yolov5Lite-s.yaml
✅ yolov5Lite-g.yaml
✅ yolov5Lite-c.yaml
✅ yolov5Lite-e.yaml
✅ yolov5-bifpn
✅ yolov5-fpn
✅ yolov5-p2
✅ yolov5-p6
✅ yolov5-p7
✅ yolov5-panet
✅ yolov5l6
✅ yolov5m6
✅ yolov5n6
✅ yolov5s6
✅ yolov5x6
✅ yolov5s-ghost
✅ yolov5-transformer 更多配置请查看 ./configs/yolo_combining 文件 ... ...

🎈Scaled_YOLOv4

✅ yolov4-p5
✅ yolov4-p6
✅ yolov4-p7 ...

🎈YOLOR

✅ yolor-csp
✅ yolor-csp-x
✅ r50-csp
✅ x50-csp
✅ yolor-d6
✅ yolor-e6
✅ yolor-p6
✅ yolor-w6
✅ yolor-ssss-dwt
✅ yolor-ssss-s2d ...

🎈YOLOv3

✅ yolov3-spp
✅ yolov3-tiny
✅ yolov3 ...

🎈YOLOv4

✅ yolov4s-mish
✅ yolov4m-mish
✅ yolov4l-mish
✅ yolov4x-mish
✅ yolov4-csp
✅ csp-p6-mish
✅ csp-p7-mish ...

🎈PicoDet

✅ PicoDet-l
✅ PicoDet-m
✅ PicoDet-s
✅ PicoDet-x ...

🚀 可选择的YOLO组合

🌟损失函数

CIoU（默认）

# 代码
python train.py --loss_category CIoU

DIoU

# 代码
python train.py --loss_category DIoU

GIoU

# 代码
python train.py --loss_category GIoU

EIoU

# 代码
python train.py --loss_category EIoU

SIoU

# 代码
python train.py --loss_category SIoU

🌟NMS

NMS（默认）

# 代码
python val.py

Merge-NMS

# 代码
python val.py --merge

Soft-NMS

# 代码
python val.py --soft

🌟多种Attention注意力机制🚀🚀🚀

具体不同注意力机制Paper以及结构图👉👉👉点击链接🔗

Attention Series🚀🚀🚀

🎈External Attention
🎈Self Attention
🎈Simplified Self Attention
🎈Squeeze-and-Excitation Attention
🎈SK Attention
🎈CBAM Attention
🎈BAM Attention
🎈ECA Attention
🎈DANet Attention
🎈Pyramid Split Attention (PSA)
🎈Efficient Multi-Head Self-Attention(EMSA)
🎈Shuffle Attention
🎈MUSE Attention
🎈SGE Attention
🎈A2 Attention
🎈AFT Attention
🎈Outlook Attention
🎈ViP Attention
🎈CoAtNet Attention
🎈HaloNet Attention
🎈Polarized Self-Attention
🎈CoTAttention
🎈Residual Attention
🎈S2 Attention
🎈GFNet Attention
🎈Triplet Attention
🎈Coordinate Attention
🎈MobileViT Attention
🎈ParNet Attention
🎈UFO Attention
🎈MobileViTv2 Attention

🌟激活函数

SiLU

# SiLU activation https://arxiv.org/pdf/1606.08415.pdf
# 代码(./utils/activations.py文件内搜索🔍)
class SiLU(nn.Module):
    ...

Hardswish

# Hard-SiLU activation
# 代码(./utils/activations.py文件内搜索🔍)
class Hardswish(nn.Module):
    ...

Mish

# Mish activation https://github.com/digantamisra98/Mish
# 代码(./utils/activations.py文件内搜索🔍)
class Mish(nn.Module):
    ...

MemoryEfficientMish

# Mish activation memory-efficient
# 代码(./utils/activations.py文件内搜索🔍)
class MemoryEfficientMish(nn.Module):
    ...

FReLU

# FReLU activation https://arxiv.org/abs/2007.11824
# 代码(./utils/activations.py文件内搜索🔍)
class FReLU(nn.Module):
    ...

AconC

r""" ACON activation (activate or not)
    AconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is a learnable parameter
    according to "Activate or Not: Learning Customized Activation" <https://arxiv.org/pdf/2009.04759.pdf>.
    """
# 代码(./utils/activations.py文件内搜索🔍)
class AconC(nn.Module):
    ...

MetaAconC

r""" ACON activation (activate or not)
    MetaAconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is generated by a small network
    according to "Activate or Not: Learning Customized Activation" <https://arxiv.org/pdf/2009.04759.pdf>.
    """
# 代码(./utils/activations.py文件内搜索🔍)
class MetaAconC(nn.Module):
    ...

🔥YOLO系列热力图可视化

🔗OD_Heatmap链接

🍋 网络模型架构图

YOLOv7🚀🎈

YOLOv6🚀🎈

YOLOX🚀🎈

YOLOv5🚀🎈

详细

YOLOR🚀🎈

PP-YOLOE🚀🎈

PP-YOLO2🚀🎈

PP-YOLO🚀🎈

Scaled_YOLOv4🚀🎈

YOLOv4🚀🎈

YOLOv3🚀🎈

以上网络模型结构图来自以下参考链接🔗 链接1 链接2 链接3 链接4 链接5 链接6 链接7 链接8

🍉 Documentation

model配置yaml文件

教程

🎓 Acknowledgement

Expand

🌰 Statement