X_LINK_ERROR when running tiny yolov4 from depthai examples

donaldlimdy

When running the yolo examples from depthai, a X_LINK_ERROR occurs immediately and the entire program could not start. Need help for this urgently. Thank you very much.

erik

Hello donaldlimdy ,
Could you share a bit more information, like depthai version, example script you are running, full log, OS, OAK model? If you are using USB, make sure to use working USB3 cable, otherwise the xlink errors are quite frequent. You could also try specifying USB2, see docs here.
Thanks, Erik

donaldlimdy

Thanks for the prompt reply, I am currently using OAK-1 with a raspberry pi 4 and tried to implement the tiny yolo v4 model obtained from the examples folder.

The script I am using is here:

The blob file is obtained from:
https://artifacts.luxonis.com/artifactory/luxonis-depthai-data-local/network/

I have tweak the code a bit for the linking part to solve the xlink error issue. The current problem I am facing is the program will run and stop immediately without prompting any errors. It seems like the line of "camRgb.preview.link(detectionNetwork.input)" is causing the problem. I am quite loss in how to solve this issue.

erik

Hello donaldlimdy , could you share the script in some other way? It gives me access denied gdrive error when I try to open it.
Thanks, Erik

donaldlimdy

Here is the link:

Thanks a lot !

erik

Hello donaldlimdy ,
From the quick glance script doesn't look incorrect. What have you changed from the TinyYolo example? Did you use the .blob (AI model) that we provide or is it custom trained?
Thanks, Erik

donaldlimdy

The TinyYolo example has a linking part of:

Linking

camRgb.preview.link(detectionNetwork.input)
if syncNN:
detectionNetwork.passthrough.link(xoutRgb.input)
else:
camRgb.preview.link(xoutRgb.input)
detectionNetwork.out.link(nnOut.input)

which lead me to a XLinkOut error, then i change it to:

camRgb.preview.link(detectionNetwork.input)

Create Outputs

xoutRgb = pipeline.createXLinkOut()
xoutRgb.setStreamName("rgb")
camRgb.preview.link(xoutRgb.input)

nnOut = pipeline.createXLinkOut()
nnOut.setStreamName("nn")
detectionNetwork.out.link(nnOut.input)

after changing the program would run but will immediately stop without error.
The blob file I used is the blob provided. Thanks a lot !

donaldlimdy

Hi I think I found the problem, it is the blob file error. Anywhere I could download a workable tiny yolo v4 trained on the MSCOCO dataset ? The blob file existing in the examples doesnt seem to be working

erik

Hello @donaldlimdy ,
I have tried the default tiny_yolo.py script with TinyYOLOv3/v4 and with syncNN set to True and False and it didn't crash in any case. What depthai version are you using?

donaldlimdy

if i not mistaken I am using the latest version of Depthai 2.15.0, is it possible to share your depthai version ? and the blob file you are using to run the tiny_yolo.py script. Thank you very much

donaldlimdy

the blob file for tiny yolo v4. Thanks a lot !

erik

Hello donaldlimdy , yes, I used depthai==2.15.0.0. I have used this Tiny-YoloV4 model. Could you try it out on your side?

donaldlimdy

I tried with depthai == 2.15.0.0, with tiny_yolo.py script and the blob file u gave me and got an error too.
The error is as follows:

donaldlimdy

I have tried for quite some time but still giving me error, any ideas how to solve ? Thanks a lot for your time.

erik

donaldlimdy Could you first try this out on a laptop, as it might be something specific to RPi/ARM?