NPU Demo and source code

CodeLogist · July 11, 2020, 7:25am

Hi everyone,

Can someone help me there on topic: Modifying NPU demo app to run on GUI desktop or support touch response in FrameBuffer mode

Its relevant to current topic only but though it would be better for new members also to create a new topic.

Thanks

alexutzu · July 11, 2020, 6:43pm

@numbqq please help me! Thanks you? demo edit in code, new camera Sony IMX586 8000 x 6000 48 MP, please help me! ISO Low Min Max 1 - 102400, Up Exposure time min max 1/25000s - 256s

RichardG · August 21, 2020, 7:02am

hi
Is there now a demo that can accept rtsp input and video file?
thanks

Electr1 · August 21, 2020, 7:20am

Hi @RichardG maybe @Archangel1235 might know something, he was talking about implementing rtsp with some demos…

Dhruv_Gaba · February 10, 2021, 3:51pm

How to run the build_vx.sh, as it gives me the following message, any help is apprecaited.
I am trying to compile my own model and build it on the khadas aml_nputy _app so as to generate binaries, but having difficulty in running this script. Please help!!!

Vladimir.v.v · February 10, 2021, 7:37pm

Hello, ask the user about this question, he seems to know the solution, good luck!

Dhruv_Gaba · February 16, 2021, 6:12am

thanks sir, take good care

Vladimir.v.v · February 17, 2021, 12:36pm

Refer also to Khadas docs.

Dhruv_Gaba · February 18, 2021, 5:20am

It’s not clear which paths to give, and it throws out the error. The steps for compiling are not very clear in the documentation. Please help me out

I have compiled my model successfully in the sdk acuity toolkit but I am not able to compile it in the aml_npu-app-master so that I could move the .so files to the board for compilation. Please guide.

Vladimir.v.v · February 18, 2021, 5:26am

Hello, Khadas team will be coming out soon after the holidays and I think they will be able to guide you right too!

Dhruv_Gaba · February 18, 2021, 7:09am

thanks will be waiting urgently

Frank · February 19, 2021, 12:43am

@Dhruv_Gaba Hello, Please download our lastest relase code , it support local compile , you don’t need to build with HOST PC

Dhruv_Gaba · February 24, 2021, 5:06pm

Thanks @Frank for the suggestion, I am able to generate the .so files now, but when I try to run the model on theVIM3 board it through’s out error saying that the model is incorrect.
Then while checking on the model I found that the website from which I took the model is different than the one is recommended in the khadas website ? the example uses 416x416 model but on the website it’s using 608x608?
any recommendations on how to get the correct model so that I can start to deploy my own models on the VIM3 board???
looking forward to your reply in anticipation

Dhruv_Gaba · February 24, 2021, 5:13pm

Also, if someone could please mentions the correct path’s and the devices where we have to do every step during the conversion specifically, that will be a great clarification too. Thanks in advance.

Dhruv_Gaba · February 24, 2021, 6:00pm

Also, how can we run the same example for a video sequence?? should we just follow the similar procedure of copying the files to detect_demo_picture to detect_demo? If you could provide some instructions on that it will be really helpful for us to move forward in the development of our product. Thanks in advance.

Frank · February 25, 2021, 1:07am

@Dhruv_Gaba Hello

When I train yolov4 , I use 416x416 . And if you get the cfg file with darknet github , it’s 608x608 , 608x608 puts too much pressure on the NPU. I do not recommend using 608x60. I have tried many versions. In the case of combining the recognition accuracy and frame rate, 416x416 is the most suitable one.

So, I suggest you train a 416x416 model yourself. When 608x608 runs on VIM3, if there are more classes recognized, the frame rate may be 0. If you need to use 608, you need to edit the source code and recompile

yolo_demo_fb_mipi/main.cpp:67:#define MODEL_WIDTH 416
yolo_demo_fb_mipi/main.cpp:68:#define MODEL_HEIGHT 416
sample_demo_fb/main.cpp:110:	amlge2d.ge2dinfo.dst_info.canvas_w = 416;
sample_demo_fb/main.cpp:111:	amlge2d.ge2dinfo.dst_info.canvas_h = 416;
yolo_demo_x11_mipi/main.cpp:67:#define MODEL_WIDTH 416
yolo_demo_x11_mipi/main.cpp:68:#define MODEL_HEIGHT 416
yolo_demo_x11_usb/main.cpp:64:#define MODEL_WIDTH 416
yolo_demo_x11_usb/main.cpp:65:#define MODEL_HEIGHT 416
yolo_demo_fb_usb/main.cpp:65:#define MODEL_WIDTH 416
yolo_demo_fb_usb/main.cpp:66:#define MODEL_HEIGHT 416
sample_demo_x11/main.cpp:107:	amlge2d.ge2dinfo.dst_info.canvas_w = 416;
sample_demo_x11/main.cpp:108:	amlge2d.ge2dinfo.dst_info.canvas_h = 416;

All need to be modified to 608

Frank · February 25, 2021, 1:09am

You can follow our docs . Here are detailed operating steps.

https://docs.khadas.com/vim3/HowToRunNPUDemo.html

https://docs.khadas.com/vim3/HowToUseNPUSDK.html

https://docs.khadas.com/vim3/HowToUseAmlNPUApp.html

https://docs.khadas.com/vim3/HowToConvertToUseNPU.html

Frank · February 25, 2021, 1:11am

There are mipi cameras or usb cameras in our demo, I have explained in the docs

Dhruv_Gaba · February 28, 2021, 2:49am

Thanks for the answer, this was the mistake I was doing. The compilation worked when I tried to compile in the VIM3 board instead of Host PC.

Dhruv_Gaba · February 28, 2021, 5:21am

@Frank I wanted to run this model but it is giving me errors during the conversion, can you please shed some light on why it is not converting to .nb format??? Thanks.