ML platform
SageMaker Ground Truth
Data labeling service with human and automated labeling workflows.
Key points
- Creates labeled datasets for machine learning training and evaluation.
- Supports human labelers through private, vendor, or public workforce options.
- Can use active learning and automated labeling to reduce manual labeling effort when appropriate.
- Supports common labeling tasks such as images, text, video, and custom workflows.
- Fits the data preparation stage before model training.
When to use it
- Choose Ground Truth when training data needs labels from humans or managed workflows.
- Use it for image bounding boxes, classification labels, text labels, or custom annotations.
- Use it when a dataset lacks the ground-truth labels needed for supervised learning.
Exam tips
- Ground Truth labels data; A2I reviews model predictions after inference.
- Ground Truth is not a model monitoring or explainability service.
- High-quality labels are critical because supervised learning depends on correct examples.
- Human review workflows can improve quality but add cost and time.