Face Recognition & People Search

For portrait photographers, wedding photographers, and event photographers, people management is often central to photo selection and delivery. Photocatalyst's face recognition runs entirely locally, automatically grouping the same person, allowing you to search, browse, and manage photos by person.

1. How Face Recognition Works

During indexing, the system automatically performs three steps:

1. Face Detection: Uses the YuNet model to scan each photo and detect the position and size of all faces
2. Feature Extraction: Extracts visual feature vectors for each detected face
3. Face Clustering: Groups faces with similar features into the same person, automatically creating person albums

The entire process runs locally — no cloud upload required.

2. Face Search

Method 1: Upload Photo Search

The most flexible search method, suitable for searching any person:

1. Click the face search icon (second from the right) on the search bar
2. Select a photo containing faces
3. The system automatically detects faces in the photo and lists all detected face thumbnails
4. Click to select the target face you want to search
5. Jump to the search results page, sorted by face similarity in descending order
6. Adjust the threshold slider to fine-tune match strictness in real time

Default threshold is 0.7. Higher threshold means stricter matching.

Method 2: Quick Search from Preview

In preview mode, interact directly with face annotation boxes on the photo:

Right-click a face box:

• Find Their Photos: Jump to all photos of that person
• Find Photos with Similar Faces: Search for other photos with similar faces based on that face's features

If the photo has multiple faces, right-clicking opens a face selector; click to choose the target face for searching.

Method 3: People Page Search

The People page lists all identified people. Click to view all photos of that person. Suitable for everyday browsing by person rather than ad-hoc searching.

3. Best Practices for Upload Photo Search

4. People Management

Naming People

1. People page → Click an unnamed person
2. Enter the person's name
3. Confirm to save

Once named, the person's name will appear in the people list across all views, making search and filtering easier.

Merging People

When the same person is mistakenly split into multiple people:

1. Select multiple people on the People page (usually different representations of the same person)
2. Use the "Merge" function
3. After confirmation, they are automatically merged — all photos belong to one person

Why do mis-splits happen?

• Photos from different age ranges (infant vs. adult)
• Different makeup/styles (everyday makeup vs. stage makeup)
• Excessively different lighting angles
• Same person with a full beard vs. clean-shaven

Lowering the face grouping threshold (Settings → AI Config) can reduce mis-splits, but may increase the risk of different people being mistakenly merged.

Splitting People

When different people are mistakenly merged into one person:

1. Enter the person's detail page
2. Find face photos that do not belong to that person
3. Select and move them to another person, or create a new person

Setting a Person Cover

1. Enter the person's detail page
2. Right-click a face photo
3. Select "Set as Cover"
4. The cover will appear on the People list page

Handling Unclassified Faces

The system displays faces that could not be matched to existing people in a centralized view:

• Manually assign to an existing person
• Create a new person based on unclassified faces
• Batch select and process multiple faces at once

5. Face Reassignment

When a face in a photo is assigned incorrectly:

1. View the photo in preview mode
2. Click the face box to select the face that needs reassignment (red highlight, supports multi-select)
3. Right-click → "Face Reassignment"
4. A dialog appears:
   • Left panel: Thumbnails of selected faces
   • Right panel: List of available target people (supports name search)
   • "Smart Face Selection" button can auto-match suggestions
5. Select the target person and confirm

6. Face Configuration Tuning

Settings → AI Config → Face Settings:

Tuning Tips:

• Many profile/small faces not detected → Lower detection threshold to 0.15
• Too many non-faces detected as false positives → Raise detection threshold to 0.3
• Enabling square crop makes the people list more uniform and aesthetically pleasing

7. Real-World People Search Scenarios

Wedding Photographer

1. Role assignment: Prepare frontal photos of the couple and guests in advance
2. Perform face search with each person's photo
3. Individual delivery: Convert search results to an album → Export for the corresponding client
4. Status tagging: Tag couple photos as "Groom"/"Bride" for cross-filtering

Event Photographer

1. Import event photos and wait for face clustering to complete
2. Check auto-clustering results on the People page
3. Name key people (guests, speakers, etc.)
4. Export by person in batches for delivery to different clients
5. Merge photos of the same person across different events (e.g., repeat clients)

Portrait/Family Photographer

1. Search all photos accumulated over the years using family members' photos
2. Cross-filter by person → timeline to find growth photos of children at each age
3. Create a "Growth Record" album arranged by year
4. Use burst picker to select the best photo from each group

Complete People Management Workflow

Indexing Complete
  ↓
People Page → View auto-clustering results
  ↓
Name each person
  ↓
Review and merge/split incorrect clusters
  ↓
Process unclassified faces (assign or create new)
  ↓
Set person covers
  ↓
Browse / Search / Export by person

Related chapters: Chapter 4: Search | Chapter 5: AI Features | Chapter 6: Multi-dimensional Browsing