face_verification 0.3.1

face_verification: ^0.3.1 copied to clipboard

• New method verifyFromImagePathIsolate() - runs verification in background isolate
• Keeps UI responsive during verification with 10-50+ registered users
• Uses BackgroundIsolateBinaryMessenger for ML Kit in isolates
• Pool-protected (max 3 concurrent operations) to prevent crashes
• Same API as verifyFromImagePath() - returns matched user ID or null

• New method verifyFromImagePathsBatch() - process multiple images safely
• Returns List<String?> with results for each image
• Automatic queuing when concurrent limit reached
• Pool protection prevents thread exhaustion crashes

• New method identifyUsersFromImagePaths() - identify users across multiple images
• Single API call for N images with configurable parallel processing
• Two-phase approach: batch face detection, then batch processing
• Returns List<ImageIdentificationResult> with per-image results
• Configurable batchSize parameter (default: 3)
• Performance: 10 images in ~5-7 seconds (vs 50s sequential)
• New ImageIdentificationResult class for type-safe results

• Extracted runModelOnPreprocessedData() for use by both normal and isolate flows
• Eliminates code duplication between verification methods
• Handles uint8/float32 tensor type conversion correctly
• Ensures identical behavior across all verification methods

• Added FaceStore.getDatabasePath() static method for isolate access
• Added ensureOpen() method to recover from closed connections
• Prevents "database_closed" errors after isolate operations

• Added pool: ^1.5.2 dependency for safe parallelism
• Pool initialized in init() with 3 concurrent operation limit
• Proper cleanup in dispose() method

• Detailed phase-by-phase timing logs for performance analysis
• Better error messages for debugging
• Graceful handling of failed operations in batch processing

Feature: Server-side embedding registration support

• New method registerFromEmbedding() - register pre-computed embeddings directly
• New method registerFromEmbeddingsBatch() - batch register multiple embeddings at once
• Enables server-side embedding generation for performance optimization
• Validates embedding size (512 dimensions required)
• Skips duplicates automatically (existing records preserved)
• Returns detailed results for tracking success/failure per embedding

Update: Upgraded tflite_flutter to 0.12.0 for improved compatibility

API: All existing methods remain backward-compatible

Feature: Group photo identification - identify ALL users in a single photo

• New method identifyAllUsersFromImagePath() - detects and identifies multiple faces in one image
• Returns List<String> of all matched user IDs (unique, deduplicated)
• Enables use cases: group attendance, event check-in, family photo tagging
• Each detected face is compared against all stored embeddings
• Works with configurable similarity threshold (default: 0.70)

Example: Updated demo app with "Identify Group Photo" feature

• New button to test multi-face identification
• Displays count and list of identified users
• Uses gallery picker for group photos

API: All existing methods remain backward-compatible

Version: Minor version bump (0.0.7 → 0.1.0) for significant new capability

Feature: Support multiple faces per user

• Users can register multiple face images using different imageId values
• Optional replace flag allows overwriting an existing (id, imageId) entry
• Verification checks all faces for a given user (staffId) or all users

Feature: Per-user face management functions

• getFacesForUser(String id) – List all face records for a user
• getFacesCountForUser(String id) – Count number of face samples for a user
• deleteFace(String id, String imageId) – Remove a specific face sample
• deleteRecord(String id) – Remove all face samples for a user

Database: Migrated schema to composite primary key (id, imageId) with createdAt timestamp

Docs: Updated README to reflect multi-face support and new management functions

Dependencies / Version: Bumped package version to 0.0.7

Feature: Added new registration status check functions

• isFaceRegistered(String id) - Check if a face is registered by ID
• isFaceRegisteredWithImageId(String id, String imageId) - Check for specific ID + imageId combination

Docs: Enhanced README with comprehensive registration management section

• Added detailed examples for registration status checks
• Included use cases for preventing duplicate registrations
• Updated feature list to highlight registration status capabilities
• Fixed iOS Podfile configuration syntax error
• Clarified imageId parameter as required (not optional)
• Added troubleshooting for "ID already exists" error

Dependencies: Updated example app dependencies

• image_picker: ^1.2.0 (from ^1.1.2)
• Updated various transitive dependencies

Dependencies: Updated plugin dependencies

• plugin_platform_interface: ^2.1.8 (from ^2.0.2)
• path_provider: ^2.1.5 (from ^2.1.4)

Docs: Complete README overhaul with improved user experience

• Added clear value proposition and use cases upfront
• Enhanced step-by-step usage guide with practical examples
• Improved code snippets with real-world scenarios
• Added comprehensive troubleshooting section
• Better organized platform setup instructions
• Added tips for optimal photo quality and threshold tuning
• Improved formatting with emojis and visual hierarchy for better readability