Failed to start job: Need data from at least two different labels to train a model. Add data with another label under "Data acquisition"

#1

hello ,
i want to train and test one classsvm sckitlearn in edge impulse. i created one processing custom block and one ml-custom bloc.
i upload my dataset and split them into train (one class) and test (two classes).
when i tried to test my model i get the following error:
Failed to start job: Need data from at least two different labels to train a model. Add data with another label under “Data acquisition”.
i tested my model (trained with one class) in colab and it worked without no problem.
why it did not worked in edge impulse.

thanks an advance

#2

Hello @aydi,

I’m not very familiar with OneClass SVM but it seems to be more like an “anomaly detection” type.

I just checked and it is not supported for Custom ML Block:

Screenshot 2025-03-19 at 17.24.16
Screenshot 2025-03-19 at 17.24.161422×1494 273 KB

I’m creating an internal feature request to see if this model type could be added as an option.

Best,

Louis

#3

thanks for your answer @louis
i bought an openmv cam h7 plus and i want to implement my code in it.
so in this case shall i wait or do you suggest any other solution in order to test my code (in colab) in openmv cam h7.
thanks an advance