Custom Labels in Projects

0

Hi I can't get my test and training images via custom labels to train. I've put in the minimum of 50% overlap. Is there something that I need to do on labels perhaps? It simply won't train. Do they all have to be exactly the same for the test and training examples? This is really frustrating and it may be that there is something wrong with my labelling protocols. Any help would be apprecaited. Thanks

asked 3 months ago127 views
1 Answer
0

Hello,

Could you provide more information on what you mean by "it won't train"? Is the training job refusing to start, or failing after starting? If there is any sort of error message, please share it as it will assist in debugging.

If the training job is refusing to start, please make sure that your datasets are properly created and labeled per our docs: https://docs.aws.amazon.com/rekognition/latest/customlabels-dg/md-create-dataset.html. Since it sounds like this could be an issue of training vs testing dataset distribution, consider training with just a training dataset as we'll split it into training and testing datasets for you (with an 80/20 train/test split).

If the job is failing after starting training, please see our docs on debugging: https://docs.aws.amazon.com/rekognition/latest/customlabels-dg/tm-debugging.html. Any service errors will need to be looked at by the Custom Labels team, but there are some common validation errors that might prevent your model from training that you would need to fix on your side.

Those are just some starting points, but more information on error messages and where you are encountering the issue will help us debug if those two points don't help unblock you.

AWS
answered 3 months ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions