張小凡

I just submitted a pull request for yolov8 example https://github.com/Tencent/ncnn/pull/5506 I hope it can be helpful to you.

> > I just submitted a pull request for yolov8 example #5506 I hope it can be helpful to you. > > Hi, I tested the yolov8.cpp example in your...

第一个问题留给其他大佬，我只能回答第二个问题， `如果target_size如果由640改为320就会导致检测结果多至4000多...` 的原因是使用ultralytics官方的model.export()函数，即便您使用了 dynamic=True 参数来导出ncnn格式的模型，仍然导出来的是固定640x640 shape，我理解这是他们的bug。 如果您确实有其他图像尺寸的识别需求的话，可能尝试下先用ultralytics官方的model.export()导出torchscript，然后再使用nihui实现的pnnx工具（https://github.com/pnnx/pnnx ）将其转换成 dynamic input shape 的ncnn模型。 估计只有这种方案才能达到您的需求，祝您好运~

> ## detail | 详细描述 | 詳細な説明 > >  > > std::shared_ptrncnn::Net ncnnNet = std::make_sharedncnn::Net(); > > 析构函数： ncnnNet->clear(); ncnnNet.reset(); ncnn::destroy_gpu_instance(); 如果有pool_allocator = new ncnn::PoolAllocator(); ，需要pool_allocator->clear();再delete。 如果有 blob_vkallocator =...

> 是的，转成wasn’t了 CPU是否有跑满？有编译好的html能否给一个让我们玩一玩？

@caoshihao996 @nihui 并没有从yolov8的ultralytics官方ort推理示例中寻找到类似您pr中所提供的根据num_anchors（或称strideNum）的值针对boundingboxs进行坐标和宽高的调整之代码，只有针对fp16和fp32的处理，如果有更多其他可供参考的文档或者Torch Yolov8后处理的代码，欢迎提出来共同学习讨论~ yolov8的ultralytics官方ort推理示例源码： https://github.com/ultralytics/ultralytics/blob/main/examples/YOLOv8-ONNXRuntime-CPP/inference.cpp#L263

Hello @Jiang-Weibo , Thank you very much for your outstanding contribution to `simpleomp.cpp`! Your modifications have given me new insights into optimization, and I have also tried applying the changes...

Hello @nihui, I found the some ncnn's example usage of OpenMP parallel sections ( `#pragma omp parallel sections` ) in the following files: ``` examples/yolov7.cpp examples/scrfd_crowdhuman.cpp examples/fasterrcnn.cpp examples/yolov5.cpp examples/yolox.cpp examples/scrfd.cpp...

> Hi there, > > Great job, and I'm looking forward to use it and compare with other frameworks like [E2FGVI](https://github.com/MCG-NKU/E2FGVI). > > Just wondering do you plan to convert...

> can you convert this project to onnx or C++ ? @lizhiqi-coder @antithing I've been trying for three working days, and I feel that the problem stems from the use...