Draw labels/faces on a kinesis stream video

To achieve real-time face detection and bounding box drawing on a Kinesis video stream, you can leverage AWS KVS in a more integrated manner. Start by ingesting your video into Kinesis Video Streams, then use Amazon Rekognition Video to perform face detection directly on the stream. Implement a serverless architecture using AWS Lambda to process the Rekognition results and generate metadata for the detected faces. This metadata can be used to draw bounding boxes on the video frames in real-time.

To accomplish this, create a Lambda function that subscribes to the Kinesis Video Stream and processes the Rekognition results. This function can then use the AWS SDK to modify the video frames, drawing bounding boxes around detected faces. The processed frames can be pushed to a new Kinesis Video Stream for output. This approach eliminates the need for separate container deployments and complex metadata storage solutions.

For distribution to multiple consumers, you can use the Kinesis Video Streams GetMedia API to allow clients to directly consume the processed video stream. If broader distribution is required, consider integrating with AWS MediaLive and AWS MediaPackage to prepare the stream for delivery via CloudFront, providing scalable and low-latency access to the processed video with drawn face bounding boxes.