BeagleY-AI still no AI working?

OK, thank you. I just ordered this debugger and look forward to a better result.

Do I need to modify the configuration file again every time I restart BeagleY-AI after a power outage? Some downloaded packages are gone. Can I have other method to save data

I cannot say exactly because I use a Yocto-based distro, not a BB distro. But I can guess two causes of your issues: filesystem corruption or data is stored on a temporary file system. In the second case there should be a script or something that can make changes permanent. I had to do that when I used the NanoBSD image.

I want to use beagley-ai to detect something.but I can’t solve problem.The following is error.I wish to get your help.
Error:(myenv) beagle@BeagleBone:~/mambaforge/object-recognition$ python3 object_detection.py --modeldir=TFLite_model
qt.qpa.xcb: could not connect to display
qt.qpa.plugin: Could not load the Qt platform plugin “xcb” in “/home/beagle/mambaforge/envs/myenv/lib/python3.9/site-packages/cv2/qt/plugins” even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.

Available platform plugins are: xcb.

Aborted

Hi, did you try running the EdgeAI app from the /opt/edgeai-gst-apps directory? If that doesn’t work with a connected display, you can try headless mode; you need to change the config file accordingly. The error message from your post indicates that your app cannot use the Qt GUI.

Since I haven’t tried it yet, I wanted to ask, can we use the CSI-2 cameras with EdgeAI App with @oleh’s instructions?

Yes, I can confirm that CSI-2 was working for me. I configured three sensors: IMX219, IMX296, and AR0234. The latter two required custom ISP configurations with de-Bayer and a Linux driver. I don’t have TI-licensed calibration software, so the colors may not be as good, but it’s fine.

1 Like

@oleh @kira9k Were you able to do the custom model compilation for this

I compiled two different YoloX models from TI but with my own data. It worked.

2 Likes

Did you use edgeai-tidl-tools or edgeai-tensorlab for the compilation. Can you share the details of compilation and deployment in this tread?

I sued the ModelMaker for edgeai-tensorlab. There is my video: https://youtu.be/uZzrB4shNyE?si=NUndUscK98T7Ehzp . I’m not sure how to translate it … sorry.

2 Likes

Thank you. I will go through this.

Hi, yes I was able to compile several models using edge ai tidl tools. I used Python API and examples from https://github.com/TexasInstruments/edgeai-tidl-tools/tree/10_00_02_00/examples/osrt_python. But not all compiled models ran successfully on the beagle

Hello, I can now call my AI accelerator (C7X) normally according to your modification, using the image of J722S (replaced). This is good news, thank you very much for sharing. But my HDMI interface connected to the display screen cannot display normally. Have you encountered this situation before? Have you tried other display output methods (for the image you changed)? At present, I tried DSI, which seems to be ineffective (overwriting the dtbo file of beagley-AI 7inch rpi, seems to have no effect)

1 Like

Hello, I haven’t used HDMI or any other display interface. All my applications were headless… Sorry, I cannot say much on this topic for now.

I followed the instructions. I have the same problems with HDMI and DSI. EDGE-AI Gallery GUI doesn’t appear with HDMI or DSI.

I get this error message:

terminate called after throwing an instance of 'std::runtime_error'
what(): No modesetting DRM card found

This errors is also specific to the DSI screen:

[   20.526350] platform bridge-regulator: deferred probe pending: platform: supplier 3-0045 not ready
[   20.535350] platform 30220000.dss: deferred probe pending: tidss: port 0 probe failed
[   20.543203] mipi-dsi 30500000.dsi.0: deferred probe pending: mipi-dsi: supplier bridge-regulator not ready
[   20.552857] platform panel0: deferred probe pending: platform: supplier 3-0045 not ready

I couldn’t get the DSI screen to work on the vanilla BY-AI either. The [RPi-7inch and 4.3inch] displays (https://www.youtube.com/watch?v=1OnKZT04JAo) don’t work…

Edit: I realizedt it. BeagleY-AI and J722SEVM use different HDMI chips.

Ukrainians are master of pattern recognition today! :smiley:

Hi! How did you realize it? Can you share some instructions?

(post deleted by author)

I read schematics. I couldn’t get it to work still.