Roboflow + AWS Rekognition Custom Labels
Label and version a dataset in Roboflow, then send it to Amazon S3 as a SageMaker Ground Truth manifest that Rekognition Custom Labels can train on.
How to connect AWS Rekognition Custom Labels to Roboflow
5 stepsLabel and version your dataset
Label your images in Roboflow, then open Versions and generate a dataset version with the train and test split you want Rekognition to use.
Add AWS keys to Roboflow
Create an IAM user that can write to your S3 bucket and create an access key for it. Paste the key under Third Party Keys in your Roboflow workspace settings.
Export to Rekognition
In the version, click Export Dataset and choose AWS Rekognition Custom Labels. Roboflow writes the dataset to S3 and gives you a train URL and a test URL.
Create the Rekognition dataset
In the Amazon Rekognition console, choose Use Custom Labels, create a project, then Create dataset. Pick "Start with a training dataset and a test dataset", choose "Import images labeled by SageMaker Ground Truth" for each, and paste the two URLs.
Train the model
Click Train model in the Rekognition project. Rekognition reports precision, recall and F1 score on the test set when training finishes.
What teams build with AWS Rekognition Custom Labels
Feed Rekognition better training data
Use Roboflow's labeling tools and augmentation to build the dataset, and keep training and hosting in your AWS account.
Compare Rekognition with a Roboflow model
Train Rekognition and a Roboflow model such as RF-DETR on the same dataset version and compare the results.
Questions
Rekognition imports labeled images from a SageMaker Ground Truth format manifest file stored in S3. Roboflow exports that format, so you do not need to write a conversion script.
Into the S3 bucket your AWS keys can write to. The export returns separate URLs for the train and test splits, which you paste into Rekognition when you create the dataset.
Yes. Roboflow reads the SageMaker Ground Truth manifest format on upload, so an existing manifest can be converted to COCO JSON, YOLO or any other format Roboflow exports.
Why are companies choosing Roboflow?
The end-to-end platform for vision AI: create datasets, train models and deploy to production.
Flexible deployment modes
Move from a scalable cloud API to edge hardware like NVIDIA Jetson without managing complex dependencies.
Integrated workflow builder
Combine multiple AI models, custom business logic and third-party integrations into one production-ready pipeline in a low-code editor.
Model monitoring and alerts
Dashboards track confidence, inference time and request volume for your deployed models, and email alerts tell your team when confidence suddenly drops.
