In partnership with MadeinWeb and AWS, Smart Fit implemented a serverless computer vision solution using Amazon Rekognition. The system automatically counts students through cameras installed in classrooms, processing and storing data in a secure and scalable way.
The processing pipeline is orchestrated by AWS Step Functions and executed on AWS Lambda, ensuring a 100% serverless architecture with automatic scalability.
The flow includes image capture, correction and pre-processing, detection via Rekognition, application of exclusion zones, multi-camera deduplication and temporal aggregation of results.
Capture cycle scheduling is managed by Amazon EventBridge, and counting results and metadata per camera/session are stored in Amazon DynamoDB.
Images and videos are stored securely in Amazon S3. Historical analyses of occupancy patterns are performed via Amazon Athena, and operational monitoring uses Amazon CloudWatch.