NO
Display methos is different in X11 demo and FB mode
For other questions, you can check the source code and the part about NPU in our document. I have a description in it.
Hey @Frank I am trying to build the application in the khadas board, but for some reason I am getting the following error? Can you tell how can I resolve it?
@Dhruv_Gaba I have test the origin aml_npu_app and aml_npu_demo_binaries, It work fine . So I think you should check youself yolov3 lib code
Ohh sorry, it was my bad I should have run the ./INSTALL first. It worked. Thanks.
Hey I was able to resolve this error by running the following commands
khadas@Khadas:~$ sudo apt update
khadas@Khadas:~$ sudo apt full-upgrade
khadas@Khadas:~$ sudo do-fenix-full-upgrade
khadas@Khadas:~$ sync
khadas@Khadas:~$ sudo reboot
and sometimes sudo apt-get update --fix-missing
it worked after that, thanks
@Dhruv_Gaba Yes, you need to use the least release OTA version , because locla build dependence not ready in old version .
got it, thanks @Frank
Thank you all for sharing your experiences here, I learned some basic things I needed to know, that i would not have otherwise.
Was looking for packages that would install needed libraries and installed libplayer-aml. This breaks normal alsa.
apt-cache show libplayer-aml
Package: libplayer-aml
Version: 2.1.0-4
Architecture: arm64
Maintainer: Khadas hello@khadas.com
Installed-Size: 1
Conflicts: aml-libamcodec-905
Replaces: libasound2
I am still looking for the required libraries to run the npu demos. Just want to warn VIM3 users - Do not install this package if you want to be able to use alsa devices normally.
I was able to partially restore normal operation with
apt purge -y libplayer-aml gstreamer-aml && apt install -y --reinstall libasound2
How to run npu demo with images comming from urls like https:// or rspt:// via x11?
@Frank the yolov3 demo code in the aml_npu_app runs on the vim NPU on the board right?? Or is it running on thr CPU of the board??
@Dhruv_Gaba All the demo is use NPU, not CPU ,but when it run it NPU , will also consume part of the CPU resources
Hi, I’m trying to convert my .cfg and .weights file following this guide: How To Transform Yolo Model | Khadas Documentation
The goal is to use the already provided yolov3 model, just test with different weights (should be pretty easy lol)
The error i’m having running 0_import_model.sh is:
./aml_npu_sdk_6.4.3/acuity-toolkit/conversion_scripts/0_import_model.sh
[…]
Traceback (most recent call last):
File “convertdarknet.py”, line 34, in
File “convertdarknet.py”, line 27, in main
File “acuitylib/vsi_nn.py”, line 142, in load_darknet
File “acuitylib/converter/convert_darknet.py”, line 525, in converte
File “acuitylib/converter/convert_darknet.py”, line 591, in build_graph
File “acuitylib/converter/convert_darknet.py”, line 426, in assign
File “<array_function internals>”, line 6, in reshape
File “numpy/core/fromnumeric.py”, line 301, in reshape
File “numpy/core/fromnumeric.py”, line 61, in _wrapfunc
ValueError: cannot reshape array of size 6519 into shape (128,256,1,1)
[30078] Failed to execute script convertdarknet
this is the .cfg i’m using
@Frank any help?
thanks!
shape 128x256x1=32768, ≠6519，your parameter setting is wrong
oh I see
what should I modify then? the script provided is a binary file and I cannot modify it, am I wrong?
thanks
@tiralonghipol yolo cfg file . Your shape setting is different from the weight file
Thanks @Frank for the clarification.
Also, is there any way to get the percentage utilization of the NPU, CPU, GPU when we run the neural network models on it??
I want to know the % utilization of the resources when I do inference of a image
@Dhruv_Gaba You can follow the NN FAQ docs
About CPU, you just need to check the CPU usage of your program
Thanks @Frank.
Do you have any information for the GPU usage during the model inference?
@Dhruv_Gaba GPU will not be used when you run NPU model .