Knowledge Base

Problem

FaceReader crashes or becomes unstable when attempting to analyze a very large number of photos (e.g., more than 12,000 images) in a single batch. RAM usage climbs to near-maximum during processing.

Details

Confirmed on machines with 16 GB RAM (reaching ~83% usage during analysis). Root cause is memory overflow from loading too many images simultaneously. Total datasets may exceed hundreds of gigabytes. Single license restricts processing to one machine.

Solution

1. Split the dataset into batches of no more than 10,000–12,000 images per batch.
2. Use the FaceReader API to automate and sequence batch processing, improving efficiency.
3. Upgrade the host machine RAM to 32 GB or higher for more stable large-batch performance.
4. Monitor memory usage during analysis and reduce batch size further if instability continues.
5. Note that a single license limits concurrent processing to one machine.