Amazon Rekognition
Rekognition is the analysis API that extracts visual information such as objects, faces, and text from images and video. It delivers common computer-vision capabilities as a callable service without requiring model training.
▶Architecture Diagram
📊 Data FlowDashed line animations indicate the flow direction of data or requests
If every text or face extraction feature starts with collecting datasets and training a vision model, that is overkill for most products. When even common visual analysis requires direct modeling work, the cost of adding one feature becomes too high.
Early computer vision projects required building both model training and inference pipelines from scratch. As demand grew for quickly adopting common visual analysis, managed APIs like Rekognition became important.
Rekognition analyzes images stored in S3 or directly provided media and returns results like objects, people, text, scenes, activities, unsafe-content signals, and face analysis or face search results. Outputs can then be post-processed with Lambda, stored in DynamoDB, or distributed through SNS.
Rekognition and SageMaker can both be used for vision tasks but serve different purposes. Rekognition provides common visual analysis features as an API, while SageMaker is better suited for building and operating custom models.
Well-suited for content tagging, face comparison, image review, post-upload auto-classification, moderation, and basic OCR-style image analysis. Not a good fit when you need to train domain-specific models yourself.