Chapter 5: AI Features

Photocatalyst integrates a local AI inference engine. All AI features run locally without network access. This chapter introduces the main AI features and how to use them.

5.1 AI Grouping

The AI Grouping page (Sidebar → AI Grouping) provides various intelligent grouping methods.

Zero-Shot Classification

Zero-shot classification is an AI classification technique that requires no training samples. Users can define custom categories, and the AI automatically assigns photos to the corresponding categories.

Adding a Single Category:

1. Click the Add button, enter the category name and description
2. The category description is used directly as the basis for search matching — the more specific, the better
3. It is recommended to write descriptions in English for more accurate classification results
4. If no description is provided, the system will use the category name for classification matching

Batch Adding Categories:

1. Click the Batch Add button
2. Enter multiple category names (separated by commas, semicolons, or line breaks)
3. You can click "Default Categories" or "Default Categories (Simple)" to quickly fill in preset category lists
4. After confirming, each category name serves as the classification basis

Running Classification:

1. Ensure at least two categories have been created
2. Click the Run Classification button, select the grouping mode (Fixed Grouping / Search Grouping)
3. Wait for the AI to complete processing
4. View photos under each category

Example:

Group by Image (Similarity)

Find all visually similar images based on selected reference photos:

1. Select one or more reference photos in the photo list
2. Right-click menu → Group by Image → Similarity Grouping
3. Adjust the similarity threshold (recommended starting at 70%)
4. Confirm and wait for the analysis to complete

Group by Image (Semantic)

Find images with similar content themes based on AI semantic understanding:

1. Select reference photos
2. Right-click menu → Group by Image → Semantic Grouping
3. Wait for the AI analysis to complete
4. The system returns semantically similar photo groups

OCR Grouping

Recognize text content in photos and group them by text keywords:

1. Select OCR Grouping on the AI Grouping page
2. Run the grouping operation
3. Wait for OCR recognition and grouping to complete
4. The system organizes photos by recognized text keywords

5.2 Face Recognition

Face recognition is one of Photocatalyst's core AI features, with all processing done locally.

Automatic Detection & Clustering

When indexing photos, the system automatically performs the following steps:

1. Face Detection: Detects all faces in photos
2. Feature Extraction: Analyzes visual features of each face
3. Face Clustering: Groups similar faces into the same person
4. Person Creation: Automatically creates albums for each person

Person Management

On the Person page (Sidebar → Persons), you can perform the following management operations:

Naming a Person:

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

Merging Persons:

1. Select multiple persons who may be the same individual
2. Use the merge function
3. The system merges them automatically after confirmation

Splitting a Person:

1. Enter the person's details
2. Select faces that do not belong to this person
3. Move to another person or create a new person

Setting Cover Photo:

1. Enter the person's details
2. Right-click on a face
3. Choose "Set as Cover"

Handling Unclassified Faces

The system lists faces that were not successfully assigned to any person:

• You can manually assign faces to existing persons
• You can create new persons based on unclassified faces
• Batch selection operations are supported

Face Reassignment

For incorrectly recognized faces, you can reassign them:

1. View photos in preview mode
2. Click on face boxes to select faces that need reassignment (multiple selection supported)
3. Right-click and choose "Face Reassignment"
4. Select the target person in the popup dialog
5. Confirm to complete the assignment

Face Search

Face interactions in preview mode:

• Single Face: Right-click face box → Search for other photos of this person
• Multiple Faces: Right-click face box → Open face selector, choose target face to search

Face Configuration

In Settings → AI Configuration, you can adjust face recognition parameters:

• Recognition Threshold: Controls the sensitivity of face detection (default 0.2)
• Face Box Scaling: Controls the face cropping range (default 1.0)
• Square Face Box: Whether to crop faces to a square

5.3 Image Quality Detection

Blur Detection

The system automatically evaluates the sharpness quality of each photo:

• Gives a quality score from 0 to 1
• The quality score is displayed in the upper left corner of the thumbnail

Score Meanings:

Hover over the icon to view the specific quality value.

5.4 AI Auto-Tagging

Tag Generation

The system automatically generates descriptive tags for photos:

• Supports both Chinese and English tags
• Covers thousands of content categories
• The tag confidence threshold can be adjusted in settings (default 0.3)

Viewing Tags

• View tags in the EXIF information of the photo preview
• Browse photos by tags on the Tags page
• Hierarchical tag management and search

Manual Tag Management

• Right-click photo → Edit → Write Tags
• Supports batch add/delete tags
• Tag search and filtering

5.5 Aesthetic Scoring & Burst Photo Selection

Aesthetic Scoring

The system automatically assigns an aesthetic score (0 ~ 10) to photos:

• Comprehensively considers composition, sharpness, exposure, and other factors
• Can be used to filter high-quality photos
• Scoring can be triggered manually

Burst Photo Selection

For burst photos, automatically picks the best one:

1. Select a group of burst photos
2. Right-click menu → Burst Photo Selection
3. The system analyzes each photo and gives a ranking
4. Choose the best photo to keep

Next: Read Chapter 6 to learn about multi-dimensional photo browsing.