Question for OAK S2

RyanLee · 2022-12-04T08:02:48+00:00

Hi, I have some question for OAK S2. As i know ir cut filter will be added in your products and i will enhance the performce of outdoor depth. I have a qu...

RyanLee

Hi,

I have some question for OAK S2.

As i know ir cut filter will be added in your products and i will enhance the performce of outdoor depth. I have a question about it. How can ir cut filter inclease its outdoor depth performance? I hope to know and check its theoretical principle about it.
to make custom model i used below link but nowadays it is not working. do you have anyother colab like for mobilenet ssd and yolo custom model? I hope to have both colab link for us.

<mobilenet ssd> but it is not working now.
https://colab.research.google.com/github/luxonis/depthai-ml-training/blob/master/colab-notebooks/Easy_Object_Detection_With_Custom_Data_Demo_Training.ipynb

Do you have any ArUco markers example which is working with oak?

Best regards,
Ryan.

erik

Hi RyanLee ,

By blocking almost all IR light sensor won't get "oversaturated" by the sun, and by but only allowing light at 940nm pass, laser dot projector should be more visible outdoors, meaning depth should be more accurate.
Could you please open an Issue on the repo, so we can track the issue? And please provide as many info as possible there, so exactly what doesn't work, full log, repro, etc.
Aruco markers aren't supported on OAK, but you can stream imgs from oak and run aruco marker detection on the host (with eg opencv).

Thanks, Erik

RyanLee

Hi Erk erik

First of all, thank you for your help always.
And i added some more information concerned with the error we met.

the code we excuted.
the error log we can see after excuting the code above.
Error message suspected as the cause of the problem from the second step.

If you need anything more, please let me know.

Best regards,
Ryan.

erik

Hi @RyanLee,
Would it be possible for you to use some more up-to-date notebooks for your object detection training, eg. Yolov4 or Yolov5? Otherwise please create github issue for tracking purposes, and we will take a look at it.
Thanks, Erik

RyanLee

Hi erik

Thank you for your quick responses. there is no problem using Yolov4 and Yolov5 for us. if it is working with oak-d-lite now. Because we will use it for our class for making custom object detection model. Could you guide us to use Yolov4 and Yolov5??

Best regards,
Ryan.

erik

Hi RyanLee ,
Please see yolov4 and yolov5 notebooks from here, they have step-by-step tutorial on training/deployment.
Thanks, Erik

RyanLee

Hi erik

Could you recommand just one link to make custom yolo model for OAK-D-Lite?? i will try that.

Best regards,
Ryan.

erik

Hi RyanLee ,
I would suggest using YoloV5:
https://github.com/luxonis/depthai-ml-training/blob/master/colab-notebooks/YoloV5_training.ipynb
Thanks, Erik