Where did it start?
Participants who receive event photos must find the shots they appear in among thousands. Organizers also need a way to collect and distribute them.
What did we build?
Organizers upload photos and share a link or QR code; participants find their photos with a selfie or bib number, no sign-up required. We implemented face-embedding search, bib OCR and scene classification ourselves, with GPU burst processing for thousands of photos at once.
- Face-embedding search from a selfie (InsightFace)
- Bib-number OCR search (RapidOCR)
- Scene classification and auto-labeling (CLIP · VLM)
- Link and QR sharing, no participant sign-up
- GPU burst processing · 5 languages · paid plans for organizers
Can this apply to my product?
Useful for photo delivery at marathons, festivals, concerts and game shows, and when planning any AI image-search product.
New projects start by defining the core features together, then a quote by scope. Payment is 50% deposit, 30% at checkpoint and 20% at handover and source delivery.
A project delivered by kinoflare. Open service
