yolov5和yolo11训练生成的检测模型均可通过ncnn转换为在android端运行的文件，查看大小，yolov5s和yolo11n的bin文件大小分别是28M和5M，感觉对yolo11精度有点不太自信了。同时部署两个版本的模型以防万一。

yolov5版本的cpp文件为yolov5_ncnn_jni.cpp，yolo11则有多个文件，初准备两版本模型用不同的cmake编译，但仅后一个能正确编译生成相应的so文件。

/*cmake {
    version "3.22.1"
    path file('libs/CMakeLists.txt')
}*/
cmake {
    version '3.22.1'
    path file('libs/y11/CMakeLists.txt')
}

比较其模型载入和检测，其主要区别在于yolov5的stride8，16，32是三个不同的值，而yolo11为一个值

ex.input("images", in_pad);
std::vector<DetectRes> proposals;
// anchor setting from yolov5/models/yolov5s.yaml
{// stride 8
    ncnn::Mat out;
    ex.extract("output", out);//output397
    ncnn::Mat anchors(6);
    anchors[0] = 10.f;
    anchors[1] = 13.f;
    anchors[2] = 16.f;
    anchors[3] = 30.f;
    anchors[4] = 33.f;
    anchors[5] = 23.f;
    std::vector<DetectRes> objects8;
    generateProposals(anchors, 8, in_pad, out, prob_threshold, objects8);
    proposals.insert(proposals.end(), objects8.begin(), objects8.end());
}
{// stride 16
    ncnn::Mat out;
    ex.extract("417", out);//781,458
    ncnn::Mat anchors(6);
    anchors[0] = 30.f;
    anchors[1] = 61.f;
    anchors[2] = 62.f;
    anchors[3] = 45.f;
    anchors[4] = 59.f;
    anchors[5] = 119.f;
    std::vector<DetectRes> objects16;
    generateProposals(anchors, 16, in_pad, out, prob_threshold, objects16);
    proposals.insert(proposals.end(), objects16.begin(), objects16.end());
}
{// stride 32
    ncnn::Mat out;
    ex.extract("437", out);//801,519
    ncnn::Mat anchors(6);
    anchors[0] = 116.f;
    anchors[1] = 90.f;
    anchors[2] = 156.f;
    anchors[3] = 198.f;
    anchors[4] = 373.f;
    anchors[5] = 326.f;
    std::vector<DetectRes> objects32;
    generateProposals(anchors, 32, in_pad, out, prob_threshold, objects32);
    proposals.insert(proposals.end(), objects32.begin(), objects32.end());
}

而yolo11则为

ex.input("in0", in_pad);
ncnn::Mat out;
ex.extract("out0", out);
std::vector<DetectRes> proposals;
generate_proposals(out, strides, in_pad, prob_threshold, proposals);

编译后运行，提示 layer 层not exist or register。

查看yolov5 param文件，新增了一层YoloV5 focus

Input            images                   0 1 images
YoloV5           focus                    1 1 images 167

重定义初始化

class YoloV5 : public ncnn::Layer {
public:
    YoloV5() {
        one_blob_only = true;
    }

    virtual int
    forward(const ncnn::Mat &bottom_blob, ncnn::Mat &top_blob, const ncnn::Option &opt) const {
        int w = bottom_blob.w;
        int h = bottom_blob.h;
        int channels = bottom_blob.c;

        int outw = w / 2;
        int outh = h / 2;
        int outc = channels * 4;

        top_blob.create(outw, outh, outc, 4u, 1, opt.blob_allocator);//.workspace_allocator);//)
        if (top_blob.empty())
            return -100;

#pragma omp parallel for num_threads(opt.num_threads)
        for (int p = 0; p < outc; p++) {
            const float *ptr = bottom_blob.channel(p % channels).row((p / channels) % 2) +
                               ((p / channels) / 2);
            float *outptr = top_blob.channel(p);
            for (int i = 0; i < outh; i++) {
                for (int j = 0; j < outw; j++) {
                    *outptr = *ptr;
                    outptr += 1;
                    ptr += 2;
                }
                ptr += w;
            }
        }
        return 0;
    }
};

DEFINE_LAYER_CREATOR(YoloV5)

并在代码中注册该层

if (yModVal == 5) {//yolo5 模型自定义层
    yolo.register_custom_layer("YoloV5", YoloV5_layer_creator);
}

成功！