- Newest
- Most votes
- Most comments
Amazon Rekognition Custom Labels does not offer a direct service feature to move or replicate a trained model from one region to another. However, you can follow a workaround to achieve this:
-
Export Training Data: Start by ensuring you have all the images and annotations that were used to train the model in US-East-2 (Ohio). This includes all positive and negative examples. If they were stored in an S3 bucket, you can easily replicate or copy them to another bucket in US-East-1.
-
Replicate Data in US-East-1: If the data is stored in an S3 bucket in US-East-2, you can use S3's cross-region replication to replicate the data to an S3 bucket in US-East-1. Alternatively, you can manually copy the data over to an S3 bucket in US-East-1.
-
Create a Custom Labels Project in US-East-1: Go to the Rekognition console in US-East-1. Create a new Custom Labels project. Begin the training process using the same dataset that you used in US-East-2.
-
Training: Unfortunately, you'll need to retrain the model in US-East-1. Although it's the same data and should yield a similar model, the training process needs to be done again as the model itself cannot be transferred across regions directly.
-
Testing: Once your model is trained in US-East-1, test it to ensure it performs as expected. If you have a specific test set that you used in US-East-2, you should use the same test set in US-East-1 to ensure consistency.
-
Production: After testing, you can then use this newly trained model in US-East-1 for your production use cases.
Relevant content
- asked 5 months ago
- AWS OFFICIALUpdated 8 months ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated 10 months ago