YOLOv11 NCNN安卓部署

YOLOv11 NCNN安卓部署

前言

yolov11 NCNN安卓部署

目前的帧率可以稳定在20帧左右，下面是这个项目的github地址：https://github.com/gaoxumustwin/ncnn-android-yolov11

上面的检测精度很低时因为这个模型只训练了5个epoch，使用3090训练一个epoch需要15分钟，后续会把训练50个epoch和100个epoch的权重更新到仓库中；

在之前复现了一个yolov8pose ncnn安卓部署的项目，在逛github的时候发现了一个关于yolov11的ncnn仓库,看了一下代码，发现作者是根据三木君大佬的代码进行改写，所以跟yolov8pose ncnn的非常的类似，所以就趁着刚改写的热乎劲，把yolov11 ncnn 安卓部署的代码改写出来；

环境配置

写这个blog的时候，安装时间为2024年11月29日

pip install ultralytics

安装后的ultralytics版本为：8.3.39，安装后的路径为:/root/miniconda3/lib/python3.8/site-packages/ultralytics

数据配置

yolov11的默认检测模型是使用COCO2017数据集进行训练，如果训练COCO数据集建议在autodl上进行训练，因为coco2017数据集在autodl上是公开数据集

如何查看autodl的共享数据

root@autodl-container-3686439328-168c7bd7:~# ls  /root/autodl-pub/
ADEChallengeData2016  COCO2017     DIV2K     ImageNet100                 VOCdevkit   mvtec_anomaly_detection.tar.xz
Aishell               CUB200-2011  DOTA      KITTI_Depth_Completion.tar  Vimeo-90k   nuScenes
BERT-Pretrain-Model   CULane       GOT10k    KITTI_Object                cifar-100
CASIAWebFace          CelebA       ImageNet  SemanticKITTI               cityscapes

数据制作

如果在实例中找到了自己需要的数据集，想使用共享数据，不能直接解压会出现只读错误，需要解压到自己的数据盘中（/root/autodl-tmp）

按照下面的流程操作即可

cd /root/autodl-tmp/
mkdir images
cd images
unzip /root/autodl-pub/COCO2017/train2017.zip
unzip /root/autodl-pub/COCO2017/val2017.zip

此时images下面只有 train2017 val2017

下载COCO2017的标签

cd /root/autodl-tmp
mkdir labels
cd labels
wget https://github.com/ultralytics/assets/releases/download/v0.0.0/coco2017labels.zip
unzip coco2017labels.zip
rm coco2017labels.zip
cd coco
rm -r annotations/
rm -r images/
rm -r LICENSE 
rm -r README.txt 
rm -r test-dev2017.txt 
rm -r train2017.txt 
rm -r val2017.txt 
mv labels/* ../
rm -r coco/  

此时labels下面只有 train2017 val2017

数据配置文件

复制COCO2017的配置文件到训练目录下

# workspace  root
mkdir train
cp /root/miniconda3/lib/python3.8/site-packages/ultralytics/cfg/datasets/coco.yaml ./train

修改coco.yaml中的path、train和val

# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: /root/autodl-tmp # dataset root dir
train: images/train2017 # train images (relative to 'path') 118287 images
val: images/val2017 # val images (relative to 'path') 5000 images

更换激活函数

YOLOv11默认使用的激活函数是SiLU,换成计算更高效的ReLU

更换激活函数后，原有的Pytorch模型需要重新训练再导出ONNX

修改/root/miniconda3/lib/python3.8/site-packages/ultralytics/nn/modules/conv.py中的第39行左右的default_act = nn.SiLU() 修改为 default_act = nn.ReLU()

训练

下载预训练权重

wget https://github.com/ultralytics/assets/releases/download/v8.3.0/yolo11n.pt

训练

训练脚本train.py

from ultralytics import YOLOmodel = YOLO('yolo11.yaml').load('yolo11n.pt')  # 加载预训练模型  还是有用的 有助于训练results = model.train(data='./coco.yaml', epochs=100, imgsz=640, batch=64, project='runs')

模型导出

模型结构修改

使用下面的方式修改模型结构不影响训练

修改/root/miniconda3/lib/python3.8/site-packages/ultralytics/nn/modules/head.py文件，修改Detect类的导出函数在其forward函数中加如下代码

if self.export or torch.onnx.is_in_onnx_export():results = self.forward_export(x)return tuple(results)

同时在Detect类新加上如下函数

def forward_export(self, x):results = []for i in range(self.nl):dfl = self.cv2[i](x[i]).permute(0, 2, 3, 1)cls = self.cv3[i](x[i]).sigmoid().permute(0, 2, 3, 1)results.append(torch.cat((dfl, cls), -1))return results

修改后的整体代码效果如下：

class Detect(nn.Module):"""YOLO Detect head for detection models."""dynamic = False  # force grid reconstructionexport = False  # export modeformat = None  # export formatend2end = False  # end2endmax_det = 300  # max_detshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initlegacy = False  # backward compatibility for v3/v5/v8/v9 modelsdef __init__(self, nc=80, ch=()):"""Initializes the YOLO detection layer with specified number of classes and channels."""super().__init__()self.nc = nc  # number of classesself.nl = len(ch)  # number of detection layersself.reg_max = 16  # DFL channels (ch[0] // 16 to scale 4/8/12/16/20 for n/s/m/l/x)self.no = nc + self.reg_max * 4  # number of outputs per anchorself.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = max((16, ch[0] // 4, self.reg_max * 4)), max(ch[0], min(self.nc, 100))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3), nn.Conv2d(c2, 4 * self.reg_max, 1)) for x in ch)self.cv3 = (nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch)if self.legacyelse nn.ModuleList(nn.Sequential(nn.Sequential(DWConv(x, x, 3), Conv(x, c3, 1)),nn.Sequential(DWConv(c3, c3, 3), Conv(c3, c3, 1)),nn.Conv2d(c3, self.nc, 1),)for x in ch))self.dfl = DFL(self.reg_max) if self.reg_max > 1 else nn.Identity()if self.end2end:self.one2one_cv2 = copy.deepcopy(self.cv2)self.one2one_cv3 = copy.deepcopy(self.cv3)def forward(self, x):"""Concatenates and returns predicted bounding boxes and class probabilities."""if self.export or torch.onnx.is_in_onnx_export():results = self.forward_export(x)return tuple(results)if self.end2end:return self.forward_end2end(x)for i in range(self.nl):x[i] = torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1)if self.training:  # Training pathreturn xy = self._inference(x)return y if self.export else (y, x)def forward_export(self, x):results = []for i in range(self.nl):dfl = self.cv2[i](x[i]).permute(0, 2, 3, 1)cls = self.cv3[i](x[i]).sigmoid().permute(0, 2, 3, 1)results.append(torch.cat((dfl, cls), -1))return results

导出的名字修改

如果需要修改输出的名称则要去修改/root/miniconda3/lib/python3.8/site-packages/ultralytics/engine/exporter.py 的 export_onnx函数

导出

导出脚本export.py

from ultralytics import YOLO# load  model
model = YOLO('best.pt')# export onnx
model.export(format='onnx', opset=11, simplify=True, dynamic=False, imgsz=640)

NCNN转化和优化

$ ./onnx2ncnn best.onnx yolov11-relu.param yolov11-relu.bin$ ./ncnnoptimize yolov11-relu.param  yolov11-relu.bin  yolov11-relu-opt.param yolov11-relu-opt.bin 1
fuse_convolution_activation /model.0/conv/Conv /model.0/act/Relu
fuse_convolution_activation /model.1/conv/Conv /model.1/act/Relu
fuse_convolution_activation /model.2/cv1/conv/Conv /model.2/cv1/act/Relu
fuse_convolution_activation /model.2/m.0/cv1/conv/Conv /model.2/m.0/cv1/act/Relu
fuse_convolution_activation /model.2/m.0/cv2/conv/Conv /model.2/m.0/cv2/act/Relu
fuse_convolution_activation /model.2/cv2/conv/Conv /model.2/cv2/act/Relu
fuse_convolution_activation /model.3/conv/Conv /model.3/act/Relu
fuse_convolution_activation /model.4/cv1/conv/Conv /model.4/cv1/act/Relu
fuse_convolution_activation /model.4/m.0/cv1/conv/Conv /model.4/m.0/cv1/act/Relu
fuse_convolution_activation /model.4/m.0/cv2/conv/Conv /model.4/m.0/cv2/act/Relu
fuse_convolution_activation /model.4/cv2/conv/Conv /model.4/cv2/act/Relu
fuse_convolution_activation /model.5/conv/Conv /model.5/act/Relu
fuse_convolution_activation /model.6/cv1/conv/Conv /model.6/cv1/act/Relu
fuse_convolution_activation /model.6/m.0/cv1/conv/Conv /model.6/m.0/cv1/act/Relu
fuse_convolution_activation /model.6/m.0/cv2/conv/Conv /model.6/m.0/cv2/act/Relu
fuse_convolution_activation /model.6/m.0/m/m.0/cv1/conv/Conv /model.6/m.0/m/m.0/cv1/act/Relu
fuse_convolution_activation /model.6/m.0/m/m.0/cv2/conv/Conv /model.6/m.0/m/m.0/cv2/act/Relu
fuse_convolution_activation /model.6/m.0/m/m.1/cv1/conv/Conv /model.6/m.0/m/m.1/cv1/act/Relu
fuse_convolution_activation /model.6/m.0/m/m.1/cv2/conv/Conv /model.6/m.0/m/m.1/cv2/act/Relu
fuse_convolution_activation /model.6/m.0/cv3/conv/Conv /model.6/m.0/cv3/act/Relu
fuse_convolution_activation /model.6/cv2/conv/Conv /model.6/cv2/act/Relu
fuse_convolution_activation /model.7/conv/Conv /model.7/act/Relu
fuse_convolution_activation /model.8/cv1/conv/Conv /model.8/cv1/act/Relu
fuse_convolution_activation /model.8/m.0/cv1/conv/Conv /model.8/m.0/cv1/act/Relu
fuse_convolution_activation /model.8/m.0/cv2/conv/Conv /model.8/m.0/cv2/act/Relu
fuse_convolution_activation /model.8/m.0/m/m.0/cv1/conv/Conv /model.8/m.0/m/m.0/cv1/act/Relu
fuse_convolution_activation /model.8/m.0/m/m.0/cv2/conv/Conv /model.8/m.0/m/m.0/cv2/act/Relu
fuse_convolution_activation /model.8/m.0/m/m.1/cv1/conv/Conv /model.8/m.0/m/m.1/cv1/act/Relu
fuse_convolution_activation /model.8/m.0/m/m.1/cv2/conv/Conv /model.8/m.0/m/m.1/cv2/act/Relu
fuse_convolution_activation /model.8/m.0/cv3/conv/Conv /model.8/m.0/cv3/act/Relu
fuse_convolution_activation /model.8/cv2/conv/Conv /model.8/cv2/act/Relu
fuse_convolution_activation /model.9/cv1/conv/Conv /model.9/cv1/act/Relu
fuse_convolution_activation /model.9/cv2/conv/Conv /model.9/cv2/act/Relu
fuse_convolution_activation /model.10/cv1/conv/Conv /model.10/cv1/act/Relu
fuse_convolution_activation /model.10/m/m.0/ffn/ffn.0/conv/Conv /model.10/m/m.0/ffn/ffn.0/act/Relu
fuse_convolution_activation /model.10/cv2/conv/Conv /model.10/cv2/act/Relu
fuse_convolution_activation /model.13/cv1/conv/Conv /model.13/cv1/act/Relu
fuse_convolution_activation /model.13/m.0/cv1/conv/Conv /model.13/m.0/cv1/act/Relu
fuse_convolution_activation /model.13/m.0/cv2/conv/Conv /model.13/m.0/cv2/act/Relu
fuse_convolution_activation /model.13/cv2/conv/Conv /model.13/cv2/act/Relu
fuse_convolution_activation /model.16/cv1/conv/Conv /model.16/cv1/act/Relu
fuse_convolution_activation /model.16/m.0/cv1/conv/Conv /model.16/m.0/cv1/act/Relu
fuse_convolution_activation /model.16/m.0/cv2/conv/Conv /model.16/m.0/cv2/act/Relu
fuse_convolution_activation /model.16/cv2/conv/Conv /model.16/cv2/act/Relu
fuse_convolution_activation /model.17/conv/Conv /model.17/act/Relu
fuse_convolution_activation /model.23/cv2.0/cv2.0.0/conv/Conv /model.23/cv2.0/cv2.0.0/act/Relu
fuse_convolution_activation /model.23/cv2.0/cv2.0.1/conv/Conv /model.23/cv2.0/cv2.0.1/act/Relu
fuse_convolution_activation /model.23/cv3.0/cv3.0.0/cv3.0.0.1/conv/Conv /model.23/cv3.0/cv3.0.0/cv3.0.0.1/act/Relu
fuse_convolution_activation /model.19/cv1/conv/Conv /model.19/cv1/act/Relu
fuse_convolution_activation /model.19/m.0/cv1/conv/Conv /model.19/m.0/cv1/act/Relu
fuse_convolution_activation /model.23/cv3.0/cv3.0.1/cv3.0.1.1/conv/Conv /model.23/cv3.0/cv3.0.1/cv3.0.1.1/act/Relu
fuse_convolution_activation /model.19/m.0/cv2/conv/Conv /model.19/m.0/cv2/act/Relu
fuse_convolution_activation /model.23/cv3.0/cv3.0.2/Conv /model.23/Sigmoid
fuse_convolution_activation /model.19/cv2/conv/Conv /model.19/cv2/act/Relu
fuse_convolution_activation /model.20/conv/Conv /model.20/act/Relu
fuse_convolution_activation /model.23/cv2.1/cv2.1.0/conv/Conv /model.23/cv2.1/cv2.1.0/act/Relu
fuse_convolution_activation /model.23/cv2.1/cv2.1.1/conv/Conv /model.23/cv2.1/cv2.1.1/act/Relu
fuse_convolution_activation /model.23/cv3.1/cv3.1.0/cv3.1.0.1/conv/Conv /model.23/cv3.1/cv3.1.0/cv3.1.0.1/act/Relu
fuse_convolution_activation /model.22/cv1/conv/Conv /model.22/cv1/act/Relu
fuse_convolution_activation /model.22/m.0/cv1/conv/Conv /model.22/m.0/cv1/act/Relu
fuse_convolution_activation /model.22/m.0/cv2/conv/Conv /model.22/m.0/cv2/act/Relu
fuse_convolution_activation /model.23/cv3.1/cv3.1.1/cv3.1.1.1/conv/Conv /model.23/cv3.1/cv3.1.1/cv3.1.1.1/act/Relu
fuse_convolution_activation /model.22/m.0/m/m.0/cv1/conv/Conv /model.22/m.0/m/m.0/cv1/act/Relu
fuse_convolution_activation /model.23/cv3.1/cv3.1.2/Conv /model.23/Sigmoid_1
fuse_convolution_activation /model.22/m.0/m/m.0/cv2/conv/Conv /model.22/m.0/m/m.0/cv2/act/Relu
fuse_convolution_activation /model.22/m.0/m/m.1/cv1/conv/Conv /model.22/m.0/m/m.1/cv1/act/Relu
fuse_convolution_activation /model.22/m.0/m/m.1/cv2/conv/Conv /model.22/m.0/m/m.1/cv2/act/Relu
fuse_convolution_activation /model.22/m.0/cv3/conv/Conv /model.22/m.0/cv3/act/Relu
fuse_convolution_activation /model.22/cv2/conv/Conv /model.22/cv2/act/Relu
fuse_convolution_activation /model.23/cv2.2/cv2.2.0/conv/Conv /model.23/cv2.2/cv2.2.0/act/Relu
fuse_convolution_activation /model.23/cv2.2/cv2.2.1/conv/Conv /model.23/cv2.2/cv2.2.1/act/Relu
fuse_convolution_activation /model.23/cv3.2/cv3.2.0/cv3.2.0.1/conv/Conv /model.23/cv3.2/cv3.2.0/cv3.2.0.1/act/Relu
fuse_convolution_activation /model.23/cv3.2/cv3.2.1/cv3.2.1.1/conv/Conv /model.23/cv3.2/cv3.2.1/cv3.2.1.1/act/Relu
fuse_convolution_activation /model.23/cv3.2/cv3.2.2/Conv /model.23/Sigmoid_2
fuse_convolutiondepthwise_activation /model.23/cv3.0/cv3.0.0/cv3.0.0.0/conv/Conv /model.23/cv3.0/cv3.0.0/cv3.0.0.0/act/Relu
fuse_convolutiondepthwise_activation /model.23/cv3.0/cv3.0.1/cv3.0.1.0/conv/Conv /model.23/cv3.0/cv3.0.1/cv3.0.1.0/act/Relu
fuse_convolutiondepthwise_activation /model.23/cv3.1/cv3.1.0/cv3.1.0.0/conv/Conv /model.23/cv3.1/cv3.1.0/cv3.1.0.0/act/Relu
fuse_convolutiondepthwise_activation /model.23/cv3.1/cv3.1.1/cv3.1.1.0/conv/Conv /model.23/cv3.1/cv3.1.1/cv3.1.1.0/act/Relu
fuse_convolutiondepthwise_activation /model.23/cv3.2/cv3.2.0/cv3.2.0.0/conv/Conv /model.23/cv3.2/cv3.2.0/cv3.2.0.0/act/Relu
fuse_convolutiondepthwise_activation /model.23/cv3.2/cv3.2.1/cv3.2.1.0/conv/Conv /model.23/cv3.2/cv3.2.1/cv3.2.1.0/act/Relu
Input layer images without shape info, shape_inference skipped
Input layer images without shape info, estimate_memory_footprint skipped

安卓代码的修改

参考这两个代码进行修改

https://github.com/gaoxumustwin/ncnn-android-yolov8-pose

https://github.com/zhouweigogogo/yolo11-ncnn

对于yolo11-ncnn有以下几个修改的地方：

1. 将softmax函数修改为了使用快速指数fast_exp的sigmoid
2. 将 cv::dnn::NMSBoxes 修改了使用纯C++代码的实现

对于ncnn-android-yolov8-pose修改为ncnn-android-yolov11主要为将各种与yolov8pose相关的内容替换为yolov11

具体的代码过程，有兴趣的可以去查看

本人技术水平不高，代码肯定还有提升优化的地方！！！

参考资料

https://github.com/gaoxumustwin/ncnn-android-yolov8-pose

https://github.com/zhouweigogogo/yolo11-ncnn

https://github.com/triple-Mu/ncnn-examples/blob/main/cpp/yolov8/src/triplemu-yolov8.cpp

https://zhuanlan.zhihu.com/p/769076635

https://blog.csdn.net/u012863603/article/details/142977809?ops_request_misc=&request_id=&biz_id=102&utm_term=yolov11%E7%9A%84%E8%BE%93%E5%87%BA%E6%98%AF%E4%BB%80%E4%B9%88&utm_medium=distribute.pc_search_result.none-task-blog-2_allsobaiduweb~default-1-142977809.142^v100pc_search_result_base2&spm=1018.2226.3001.4187