Rekognition Face to UserId association limit.

The 100-face limit per UserID in Amazon Rekognition is indeed a constraint that can be challenging for use cases involving personal photo collections where the same individuals appear frequently. This limit is not adjustable through the standard Service Quotas console, as it's a fundamental design aspect of the service rather than a configurable quota.

The UserID feature in Rekognition is primarily designed for scenarios where you need to identify a relatively small number of distinct faces for each user, such as employee authentication or customer identification systems. It's not optimized for personal photo collections where the same individuals appear many times.

For your use case of indexing 100k - 1M personal photos with frequent appearances of the same individuals, you might need to consider alternative approaches:

1. Multiple Users per Person: You could create multiple UserIDs for each real person, effectively distributing their faces across these IDs. This would require additional management on your end to track which UserIDs belong to which actual person.

2. Face Grouping: Instead of relying solely on Rekognition's UserID system, you could implement your own face grouping logic. Use Rekognition to detect and compare faces, but manage the associations in your own database. This gives you more flexibility in handling multiple appearances of the same person.

3. Selective Indexing: You could selectively index faces, choosing representative samples for each person rather than indexing every single appearance. This might work well for personal collections where you don't need to identify every single instance of a person.

4. Multiple Collections: Consider using multiple face collections, distributing faces across them to avoid hitting the limit in a single collection.

5. Custom Solution: For large-scale personal photo organization, you might need to build a custom solution that uses Rekognition's face detection and comparison capabilities, but implements its own logic for grouping and managing identities.

If none of these approaches suit your needs, you may want to contact AWS Support to discuss your specific use case. While the 100-face per UserID limit is not typically adjustable, they might be able to provide additional guidance or suggest alternative solutions within the AWS ecosystem for your particular scenario.
Sources
ServiceQuotaExceededException - AssociateFaces | AWS re:Post
Searching faces in a collection - Amazon Rekognition