yolov5_obb icon indicating copy to clipboard operation
yolov5_obb copied to clipboard

Run yolov5_obb on a Smartphone ?

Open Avant-Gardiste opened this issue 3 years ago • 14 comments

Hi, I was wondering if it might be possible to run your model even on the limited computing power of a smartphone. Before diving too deep into it, I wanted to get the opinion of more experienced people in the field.

If you think it is possible, is a rough estimation for fps possible ?

If you think it is not possible could you briefly elaborate on the reasons please ?

Any insights are appreciated. Thank you very much for your work!

Avant-Gardiste avatar May 24 '22 09:05 Avant-Gardiste

@hukaixuan19970627 Could you please answer ?

Avant-Gardiste avatar May 25 '22 07:05 Avant-Gardiste

I'm sure it will work on smartphone because yolov5s_obb has low computing power requirements. image But before that you'd better convert the CUDA code of the obb_nms function to python/C++ code.

hukaixuan19970627 avatar May 25 '22 12:05 hukaixuan19970627

@hukaixuan19970627 any idea how can i convert obb_nms funtion to python/C++ ?? Thank you so much for your responses !

Avant-Gardiste avatar May 25 '22 13:05 Avant-Gardiste

Sorry, I don't have any idea, may be you can refer to this repo. Please let me know if you run yolov5_obb on smartphone successfully.

hukaixuan19970627 avatar May 25 '22 13:05 hukaixuan19970627

@hukaixuan19970627 Yep, I'll keep you in touch ! Thk you ;)

Avant-Gardiste avatar May 25 '22 13:05 Avant-Gardiste

Hi @hukaixuan19970627, I want to inform you that I had successfully implemented the model to an Android device and the frame rate in real time is slightly good ! In order to had this done, i implemented the SOFT_NMS function in Java and replace it whit traditional NMS. I don't know how this could be done without your work. So, thank you so much for you effort and your contribution to the open source community !

Below is a capture that shows inference on smartphone ! Screenshot_20220711-091831

Avant-Gardiste avatar Jul 11 '22 07:07 Avant-Gardiste

bravo!!!

hukaixuan19970627 avatar Jul 17 '22 10:07 hukaixuan19970627

Great work @Avant-Gardiste !! I already ran yolov5 model on android but I'm also trying yolo_obb and running on android device.It would be great if you help me in this regards. Thank you in advance.

apanand14 avatar Aug 16 '22 07:08 apanand14

@apanand14 Sure, gladly. Ping me whenever you want. Here is my LinkedIn profile : https://www.linkedin.com/in/el-mehdi-tafik-740043202/

Avant-Gardiste avatar Aug 16 '22 14:08 Avant-Gardiste

Thank you so much for sharing your profile but I'm not using it. Would you mind if I contact you on your email? @Avant-Gardiste

apanand14 avatar Aug 17 '22 06:08 apanand14

Sure, no pb !

Avant-Gardiste avatar Aug 17 '22 07:08 Avant-Gardiste

can you share your email please?

apanand14 avatar Aug 17 '22 08:08 apanand14

@apanand14 [email protected]

Avant-Gardiste avatar Aug 17 '22 08:08 Avant-Gardiste

@Avant-Gardiste : would it be possible if you could share your implementation? Thanks.

nik13 avatar Aug 26 '22 19:08 nik13

did anyone run yolo_obb with ncnn framework on android?

apanand14 avatar Oct 10 '22 14:10 apanand14

hf62580 avatar Dec 23 '22 06:12 hf62580

有人在 android 上用 ncnn 框架运行 yolo_obb 吗?

hf62580 avatar Dec 23 '22 06:12 hf62580

Hi!! great! thank you for your response. would you like to guide for the deployment or share inference code if you do not mind? thank you in advance

apanand14 avatar Feb 03 '23 14:02 apanand14

hi if you want to run on cpu you just have to download python file softnms_rotate.py from [https://github.com/ZhangGongjie/Soft-NMS-for-Rotated-Rectangles] and put that file into utils folder. then go to general.py and do these 2 step. it will work on cpu as well as other devices step_1 step_2

ibtehajali67 avatar Mar 07 '23 11:03 ibtehajali67