weida wang

Results 65 comments of weida wang
trafficstars

@nmss I just tried disable the setting `markdown-toc.updateOnSave`, but the extension still automatically restore content to original one

@wczqhd Do you solve the problem now ? If not, could you please provide deeper infomations like what is the final line of code in inference_talkinghead() that throws this error...

Hi, bro, i am doing exactly the same thing, how is it going on now ?

This device hikey970 with usb camera also have problem with CameraX used by [original nanodet ncnn android app](https://github.com/cmdbug/YOLOv5_NCNN), it will crash when it try to open camera with error :...

> nihui 你好,我尝试使用了yolov5ncnn,检测精度相比其他yolov5的Android实现都要高,但是我尝试进行摄像头实时检测,FPS不到5帧,和其他开源的yolov5项目速度差不多。 > > 但是YOLOv5作者开发的IOS端应用iDetection也是用YOLOv5进行摄像头实时检测,其FPS在30左右,有几个问题我不太明白。 > > 一、请问为什么在Android和IOS上YOLOv5的检测速度会相差这么多? > 二、我发现基于NCNN进行YOLOv5检测时,分别从YOLOv5的三个输出层提取结果是最耗时间的(代码为ex.extract("output", out);),请问有什么方法能够对它进行优化加快检测速度吗? > > 期待nihui的解答! nihui未到,我先来了... 1. 我的ipad Air2上面运行yolov5s的帧率也才4.4FPS而已,竟然跟iphone差别这么大? ![image](https://user-images.githubusercontent.com/43233772/98440416-7d0d3b80-2133-11eb-98e0-44d35adb1388.png) 2.大佬能提供一下yolov5 ncnn android实时监测的app吗,nihui这个是检测图片嘛,感觉检测视频更实用

谢谢分享,我也发现nihui的这个yolov5比[另一个yolov5 android](https://github.com/cmdbug/YOLOv5_NCNN)要高,速度的话你发的这个我看不出FPS多少,但卡顿情况目测起来差不多,所以我也很好奇nihui的这个为什么比cmdbug的效果好很多,而且我发现nihui的yolov5s.bin大小才14M多,cmdbug的yolov5s.bin有27M多。我还没深入研究两者的差别,大佬@cmdbug ,您有空瞧一瞧吗? ![image](https://user-images.githubusercontent.com/43233772/98459002-471e9480-21d1-11eb-9329-bcb57a5426d6.png) ![image](https://user-images.githubusercontent.com/43233772/98459005-4ab21b80-21d1-11eb-89ac-52292ff8ab25.png)

I am curious about this too, do you mean just replace with yolo-fastest bin and param ?

I think i find my answer from [here](https://github.com/KinWaiCheuk/nnAudio).

我试了一下,确实有这个问题: 本仓库app效果: ![6E477AC0DF16BC0FF54BE9F92B6E65E1](https://user-images.githubusercontent.com/43233772/166134771-9bd7f8d8-eb0f-4c24-bf6c-cb0d4b62e537.jpg) mediapipe官方app效果: ![D16FC0AC844440F1A56A0725A9B18292](https://user-images.githubusercontent.com/43233772/166134747-9bf275b2-2c9d-4465-8c64-e450f9905103.jpg) 测试图片: ![face](https://user-images.githubusercontent.com/43233772/166134678-8e8b3b64-c5c2-4d3f-9625-462713aa343d.jpg)