Help us improve the AWS re:Post Knowledge Center by sharing your feedback in a brief survey. Your input can influence how we create and update our content to better support your AWS journey.
Unable to train custom label object localization model: Amazon Rekognition experienced a service issue
Hello,
I have tried to train Amazon Rekognition Custom Label object localization model on 105,000 images. I have tried training the model 2 times and reduced training datasets to 70,000 and 55.000, but all I am getting a "training failed" message after almost 48 hours of training with status message: "Amazon Rekognition experienced a service issue." The model seemed can successfully trained when I used a smaller subset of the data with only about 60 labeled images. Is the dataset size the issue? I found in web, "AWS Rekognition CustomLabels supports upto 250,00 images for Object Detection/Localization." I was also able to train image classification model use Amazon Rekognition Custom Label on 105,000 images successfully with no issue.
AWS Region in which I ran these trainings: us-west-2 AWS account: XXXX-XXXX-5633 Approximate time of trainings: 12/22/2022-12/28/2022, with failed model name list:
- *****.2022-12-27T19.35.49
- *****.2022-12-26T16.04.56
- *****.2022-12-25T16.19.44
- *****.2022-12-22T20.43.45
Any help would be appreciated! Thank you.
- Topics
- Machine Learning & AI
- Language
- English
- Newest
- Most votes
- Most comments
Thank you for reaching out to us. We identified and fixed the underlying issue, and you can now retry your training. Please feel free to create a support case to us directly for further support.
Relevant content
- asked 2 years ago
- asked 2 years ago
- AWS OFFICIALUpdated 9 months ago
I retried training but still failed with "Amazon Rekognition experienced a service issue".