Separation of Event Photos for Graduation Ceremonies Using Machine Learning with AWS Rekognition
Medium
Photo Printing
Brazil
Medium
Brazil
Photo Printing
Pixelhouse is a leading company in large-scale photo printing, located in Barueri, São Paulo. With the capacity to produce over 500,000 photographs and thousands of customized albums per day, Pixelhouse offers innovative solutions for graduation event management companies.
Since 2019, Pixelhouse has been providing a SaaS platform for graduation event management, initially covering only the layout and printing of custom albums. However, by adopting Machine Learning technologies, Pixelhouse expanded its platform to include automatic separation of graduation photos, allowing companies to produce customized albums with unique photos for each student quickly and efficiently.
The Challenge of Event Management Companies
The management of graduation events presents considerable challenges for the companies involved, including the capture and storage of large volumes of photographs of graduates, taken throughout various stages of this journey. In certain situations, this scenario can involve more than 100,000 photographs, accumulating several terabytes of data stored for a single event.
Historically, companies have resorted to inadequate tools to separate the photographs, resulting in fragile backups susceptible to data loss or corruption. Additionally, the process of uploading the photographs to Pixelhouse's platform for later printing often led to rework and operational inefficiencies.
Solution
Pixelhouse and CloudDog have developed an application to integrate into their existing SaaS, streamlining the photo management process at graduation events. The solution includes long-term storage, photo separation, digital album delivery, and integration with the layout and printing system.
Event management companies can now upload photos directly to Amazon S3 Buckets, which automatically initiates the separation process using AWS Rekognition. Then, heuristics are applied to classify the type of photo and optimize manual review.
The architecture was designed to handle peak processing during graduation events, which mainly occur in December and January. For this purpose, serverless technology services such as Amazon CloudFront, API Gateway, DynamoDB, Step Functions, Lambda, Simple Notification Service, and Simple Queue Service were used.
Processing is done asynchronously and in parallel, which can result in up to 800 lambdas being executed simultaneously, optimizing images at different resolutions for later use in the workflow.
Solution Architecture Diagram
The solution architecture diagram illustrates how CloudDog and Pixelhouse overcome technology challenges using serverless services provided by Amazon Web Services.
Results
With the new application, graduation event management companies can now upload event photos in an optimized manner and receive digital files by email within 24 hours after the event.
The accuracy rate of AWS Rekognition has proven to be quite high, exceeding 99% in most scenarios, even in photos with faces covered by masks, which was common during the COVID-19 pandemic.
Now, Pixelhouse has a virtually unlimited platform capacity to scale the graduation album production process for all graduation event companies in Brazil without worrying about storage or processing capacity.
This project underwent review by Amazon Web Services architects and was used as a reference for validating our expertise in serverless application development, including Amazon API Gateway, AWS Lambda, and Amazon DynamoDB.
AWS Partner Summit SP 2023
On the second day of AWS Partner Summit SP 2023, the success story of PixelHouse, a company with a SaaS for graduation event companies facing challenges in storage, separation, layout, and printing of graduation photos, was presented.
The development of the application that integrated with Amazon S3, API Gateway, Step Functions, Lambda, DynamoDB, and AWS Rekognition for photo separation and optimization was showcased.
With the new application, graduation event management companies can upload event photos in an optimized manner and receive digital files by email within 24 hours. The accuracy rate of AWS Rekognition was above 99%.
Now, Pixelhouse's platform has virtually unlimited capacity to scale the graduation album production process.
