Face Recognition & Matching
The AI Photo Gallery’s core feature is face recognition — the system detects every face in your event photos, groups them by person, and lets attendees find all their photos with a single selfie.
How Face Recognition Works
Section titled “How Face Recognition Works”- Detection — AI scans each uploaded photo and identifies individual faces.
- Grouping — Detected faces are compared and clustered. Faces that look like the same person are placed in the same group.
- Matching — when an attendee takes a selfie, the system compares it against all face groups.
- Results — the attendee sees every photo where their face appears.
Face Recognition Stats
Section titled “Face Recognition Stats”Go to AI Gallery → Dashboard to see face recognition metrics:
| Metric | What It Shows |
|---|---|
| Total Groups | Number of unique people detected across all photos |
| Matched Groups | Groups where the person has been identified (selfie matched) |
| Unmatched Groups | Groups where the person hasn’t been identified yet |
| Total Faces | Total individual face detections across all photos |
| Average Photos Per Person | How many photos the average attendee appears in |
| Match Rate | Percentage of groups that have been successfully matched |
Managing Face Groups
Section titled “Managing Face Groups”Go to AI Gallery → Face Groups (or the face management section):
View Face Groups
Section titled “View Face Groups”Each group shows:
- Thumbnail — a representative face from the group
- Photo count — how many photos this person appears in
- Status — matched (identified) or unmatched
- Attendee name — if matched, shows the attendee’s registration name
Merge Duplicate Groups
Section titled “Merge Duplicate Groups”Sometimes the AI creates two groups for the same person (e.g., different lighting or angles). To merge:
- Select the duplicate groups.
- Click Merge.
- All photos from both groups are combined under one identity.
Split Incorrect Groups
Section titled “Split Incorrect Groups”If the AI grouped two different people together:
- Open the group.
- Select the incorrectly grouped photos.
- Click Split to create a new group from the selected photos.
Selfie-Based Photo Retrieval
Section titled “Selfie-Based Photo Retrieval”Attendees find their photos using a selfie:
- Attendee opens the AI Photo Gallery link (from email, WhatsApp, or event website).
- They take a selfie or upload a photo of themselves.
- The AI matches their face against all detected groups.
- They instantly see every photo they appear in.
- They can download, share, or save their photos.
Selfie Registration
Section titled “Selfie Registration”For events that collect attendee photos during registration:
- The system can automatically match registration photos against face groups.
- Attendees don’t even need to take a selfie — their photos are ready when they visit the gallery.
- This works best when your registration form includes a photo upload field.
Photo Attendee Management
Section titled “Photo Attendee Management”For each photo, you can see which attendees appear in it:
- Open a photo.
- View the Faces Detected panel — shows all faces found in this photo.
- Each face links to its group (and to the matched attendee, if identified).
This is useful for:
- Verifying that key speakers and VIPs were properly detected
- Checking that group photos correctly identify all participants
- Quality-checking the AI’s face detection accuracy
Improving Recognition Accuracy
Section titled “Improving Recognition Accuracy”Face recognition accuracy depends on photo quality:
| Factor | Impact |
|---|---|
| Lighting | Well-lit photos have much higher detection rates |
| Face angle | Front-facing faces match best; extreme profiles may be missed |
| Resolution | Higher resolution = better face detection |
| Occlusion | Masks, sunglasses, or hands covering faces reduce accuracy |
| Distance | Faces that are too small in the frame (far from camera) may not be detected |
Best Practices
Section titled “Best Practices”- Use professional photographers — well-lit, high-resolution photos dramatically improve face detection accuracy.
- Capture front-facing photos — instruct photographers to capture people facing the camera when possible, not just candid side angles.
- Review face groups — spend 10–15 minutes after processing to merge obvious duplicate groups and split any incorrect groupings.
- Promote selfie retrieval — the gallery is only valuable if attendees use it. Send a WhatsApp campaign with the gallery link right after the event.
- Check VIP coverage — verify that your keynote speakers, sponsors, and VIP guests are correctly detected and grouped. Manually merge groups if needed.