Hi I'm newbie intern in computer vision domain I have trained my model of Anomaly detection on my personal machine now I want to deploy my model on OAK-D device. Can anyone help me with it

Hello @"Areeba"#253, that's awesome to hear! We have a tutorial on how to convert trained models into IR/blob for deployment to the DepthAI [here](https://docs.luxonis.com/en/latest/pages/model_conversion/). There are tons of useful resources linked from that page that can help you :) Thanks, Erik

@"erik"#p1131 THANKS Erik Yes I've followed the steps to create blob file from the link you shared above, but the issue I'm having is how to use that blob file to deploy model on OAK-D. If you could refer to any specific tutorial or example that would be really helpful. Thanks again! :)

Hello @"Areeba"#253, So you have to load your blob into the neural network node and later get the result of the NN. [This demo](https://github.com/luxonis/depthai-experiments/blob/master/gen2-age-gender/main.py) does a similar thing - [loading a blob](https://github.com/luxonis/depthai-experiments/blob/master/gen2-age-gender/main.py#L54-L57) and [decoding the result](https://github.com/luxonis/depthai-experiments/blob/master/gen2-age-gender/main.py#L139-L150). Thanks, Erik

Hi @"erik"#175 , I've followed your steps of creating blob file and creating source code file as shown in demo, but I am getting this error [URL=https://i.imgur.com/Wj9WdJQ.png][IMG]https://i.imgur.com/Wj9WdJQh.png[/IMG][/URL] Here's the g-drive link of my blob file, source file and colab notebook which I used to create blob file https://drive.google.com/drive/folders/1ZsJbqNCPYI4TLmtF1_ydCoJnqEd7Xp11?usp=sharing (I am running this example of Mask detector from example demo link you shared above) THANKS, Areeba

Hello @"Areeba"#253 , I believe the issue is with the NN result decoding. As the error states, size of the bounding boxes array is 0, and you can't access the first element. Different NN models can have a different output so you have to change the decoding part of the code. Here are some examples of decoding the NN result, just to give you an idea: - [object detection (efficientDet)](https://github.com/luxonis/depthai-experiments/blob/master/gen2-efficientDet/main.py#L65-L83) - [classification (fire detection)](https://github.com/luxonis/depthai-experiments/blob/master/gen2-fire-detection/main.py#L257-L269) - [semantic segmentation (deeplab)](https://github.com/luxonis/depthai-experiments/blob/master/gen2-deeplabv3_depth/main.py#L149-L156) Thanks, Erik

Thanks @"erik"#p1137 Yes the this issue is resolved now. I've created .blob file by using this [colab notebook](https://colab.research.google.com/github/luxonis/depthai-ml-training/blob/master/colab-notebooks/Easy_TinyYolov3_Object_Detector_Training_on_Custom_Data.ipynb) I'm using this tutorial to run .blob file in depthai [URL=https://i.imgur.com/ndkMrys.png][IMG]https://i.imgur.com/ndkMrysh.png[/IMG][/URL] But its showing this error[URL=https://i.imgur.com/ODkO3b7.png][IMG]https://i.imgur.com/ODkO3b7h.png[/IMG][/URL] Thanks Areeba

Hello @"Areeba"#253 , so you used the `tiny yolo v3 decoding` example, modified parts that are specified by the 3 bullet points and it returns you the `XlinkOut 220` error? Or is it resolved already? I tried those steps myself (when writing the instructions) and it worked flawlessly, sorry for the troubles. Thanks, Erik

Yes @"erik"#175 I've modified those three bullet points but its still showing the error I've even made changes according to [this ](https://github.sre.pub/luxonis/depthai-ml-training/issues/4) Thanks Areeba

@"Areeba"#253 could you please share the trained blob that you got from the notebook? I would love to try it out - and sorry for the inconvenience, if there is an issue with the notebook I will fix it as soon as we figure out what's wrong. Thanks, Erik

[here](https://drive.google.com/drive/folders/1ZsJbqNCPYI4TLmtF1_ydCoJnqEd7Xp11?usp=sharing) is the g-drive link

Deploy pretrained model on OAK-D device

erik

Areeba Sorry for the trouble, but could you share your blob? I would like to try it out, just so we can cross out one possibility for the issue you are having.

Areeba

erik Here's the blob file

Areeba

Here's the blob file

erik

Hello @Areeba , sorry for the delay and inconvenience, I can confirm that there's a problem with the blob and we will check the notebook & integration of the blob with depthai tomorrow.

erik

Hello @Areeba , I'm unsure why it wasn't working for you, but after I went step by step through the yolov3 tutorial and it worked as expected. It might be you had an old tutorial. Script code here, blob here. Run with python3 mask-nomask-yolov3.py yolov3masks.blob 🙂

Areeba

https://colab.research.google.com/github/luxonis/depthai-ml-training/blob/master/colab-notebooks/Easy_TinyYolov3_Object_Detector_Training_on_Custom_Data.ipynb
Is this the updated one?

erik

Areeba yes, that is the one I used to train the blob I shared above🙂

Mauricio

Hi, I'm a newbie with the OAK-D.

I wanted to let you know that the shared Colab is not working anymore with the tf 1. X versions because now Colab works with Python 3.10. Anyways, Roboflow's examples have the same problem.

erik

Hi Mauricio ,
Which notebook did you try? The latest Yolov5/6/7/8?

Mauricio

I apologize, I did not know about that repo. I checked it and it solves my FP16 troubles and more. I'll try them.

Thank you!!!

Mauricio

Excuse me, I wonder if it might be possible to run also segmentation models if I change the training to segment. Or I'm just going to lose my time 🤔

jakaskerl

Hi Mauricio
With YOLO 5 and 6 it should work, 7 or 8 are currently detection only i think.

Thanks,
Jaka

Mauricio

jakaskerl Thank you!!

« Previous Page