Hi everyone, we have been trying to create a classifier neural network to detect the orientation of a cone using limelight. We have been able to successfully create the model on teachable machine but when we try to convert the model it fails and says “Something went wrong while converting.” The model has 10 classes for different orientations of the cone. When we tried to build the model with 5 classes it succeeded, but we are hoping to be able to have the model with 10 so we can get more detailed information. Any solutions or ideas are greatly appreciated! Thank you!
I don’t have any specific ideas, but does it matter which 5 classes you export? I’m wondering if it is having trouble with a particular class, or if it is hitting some other limit. Roughly how many images are in your training data set?
I don’t think it matters which 5 we export as we have tried a couple variations. We have also tried different numbers of images in the classes. The first time we did 144 per class and when that didn’t work we decreased it. We tried with a few different numbers per class all the way down to 48 images per class but never got it to work. I think that the problem is something on their end related to the number of classes, but am not sure how to fix that as we would like to have all of the classes if possible. I am going to try to contact them and hopefully they respond.
Any updates on this? I tried also to convert a tflite quantized model but I’m stuck at something went wrong while converting. I have 10 classes with 300 images each class
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.