search

dog-qiuqiu / Yolo-Fastest

:zap: Based on yolo's ultra-lightweight universal target detection algorithm, the calculation amount is only 250mflops, the ncnn model size is only 666kb, the Raspberry Pi 3b can run up to 15fps+, and the mobile terminal can run up to 178fps+
Other
1.97k stars 428 forks source link

operating speed of Raspberry Pi #29

Open 2684160504 opened 3 years ago

2684160504 commented 3 years ago

Hello, I am on 1660TiGPU in ubuntu16.04, yolofastest speed is 30 frames, and about 1 frame on Raspberry Pi. Is this normal?Looking forward to your reply, thank you.

liuyijian commented 3 years ago

I have tried yolo-fastest at RaspberryPi 4 (8GB) with opencv 4.4，and the result is about 7~8 fps.

2684160504 commented 3 years ago

I have tried yolo-fastest at RaspberryPi 4 (8GB) with opencv 4.4，and the result is about 7~8 fps.

OK,Thank you very much.

sudo-install-MW commented 3 years ago

Hi @liuyijian can you please provide some information on what is the input resolution you had your model tested for in raspberry pi?

liuyijian commented 3 years ago

Hi @sudo-install-MW My input resolution to the model is 320*320, which is the default setting in yolo-fastest.cfg. I use function cv2.dnn.blobFromImage(...) to transform the original frame (1280*720) that capture from the camera, to 320*320 as model input. As I am not sure if I can change the default width and height paremeter in yolo-fastest.cfg, I make temporary practice like this. Hope to learn some practical trick from you.

jereter commented 3 years ago

I have tried yolo-fastest at RaspberryPi 4 (8GB) with opencv 4.4，and the result is about 7~8 fps.

兄弟，树莓派4上面执行ncnn_sample里面的指令，一直报错benchmark.h文件找不到，有遇到过吗

liuyijian commented 3 years ago

@jereter ncnn_sample里的前面几个include需要的文件在Tencent/NCNN仓库的src目录里，具体地址如下： https://github.com/Tencent/ncnn/tree/b93775a27273618501a15a235355738cda102a38/src