SmartJavaAI Documentation

SmartJavaAI is a lightweight, free, and offline AI toolbox designed specifically for Java developers. It aims to provide a zero-threshold experience for integrating AI algorithms into Java projects, following the philosophy of being 'as easy to use as Hutool'.

Key features include:

• Zero-threshold integration: Call complex functions like face recognition, object detection, and OCR without needing to understand underlying AI implementations.
• Multi-framework support: Leverages DJL (Deep Java Library) and JNI-based C++/Python algorithms to support PyTorch, TensorFlow, ONNX, and Paddle.
• Offline capabilities: Designed for offline use, making it suitable for privacy-sensitive or disconnected environments.
• Android Support: Offers a commercial Android SDK for mobile-specific offline face recognition needs.

The SmartJavaAI Android version is a high-performance, fully offline facial recognition solution designed for mobile devices. It does not require an internet connection to function. The SDK consists of four core modules:

1. Face Detection: Locating faces within an image or video stream.
2. Face Registration: Enrolling new faces into the system.
3. Face Comparison (1:1): Verifying if two faces belong to the same person.
4. Face Search (1:N): Searching for a specific face within a database of multiple faces.

The translation example is organized as follows:

└── main
    ├── java
    │   └── smartai
    │       └── examples
    │           └── nlp
    │               └── translation 机器翻译
    │                   └── TranslationDemo.java 
    └── resources
        └── logback.xml

Core logic resides in TranslationDemo.java, and logging configuration is handled by logback.xml in the resources folder.

The speech-examples project contains several implementations for speech processing:

• ASR (Automatic Speech Recognition):
  • SherpaAsrDemo.java: Demonstrates speech recognition using the sherpa-onnx engine.
  • SpeechRecognizeDemo.java: Demonstrates speech recognition using other available models.
• TTS (Text-to-Speech):
  • TtsDemo.java: Demonstrates speech synthesis (converting text to speech).

Detailed usage instructions and functional descriptions can be found in the comments at the top of each Java file.

The examples directory contains several standalone Maven projects demonstrating specific capabilities of the SmartJavaAI library. You can use these as templates for your own implementation:

• face-example: Demonstrates face detection and face recognition.
• vision-example: Demonstrates general computer vision tasks including object detection, object segmentation, and image classification.
• ocr-example: Demonstrates Optical Character Recognition (OCR), text recognition, and license plate recognition.
• translate-example: Demonstrates machine translation capabilities.
• speech-example: Demonstrates speech recognition (STT) and speech synthesis (TTS).

The examples directory provides the following functional modules:

• face-example: Face detection, face recognition, etc.
• vision-example: General vision tasks like object detection, segmentation, and image classification.
• ocr-example: OCR text recognition, license plate recognition, etc.
• translate-example: Machine translation.
• speech-example: Speech recognition (ASR) and speech synthesis (TTS).

The face-example project provides Java implementation examples for various facial analysis scenarios using the SmartJavaAI SDK. Key capabilities include:

• Face Detection (facedet): Detects faces in images or video streams and provides bounding boxes.
• Face Recognition (facerec): Supports 1:1 verification and 1:N searching by extracting facial features and managing a face library.
• Liveness Detection (liveness): Determines if a face is a real person using static images or camera video streams.
• Face Attribute Detection (attribute): Detects attributes such as gender and age.
• Expression Recognition (expression): Recognizes 7 types of expressions (e.g., neutral, happy, sad).
• Face Quality Assessment (quality): Evaluates image quality based on brightness, clarity, completeness, pose, and resolution.

SmartJavaAI is organized into modular components. You can include specific modules based on your requirements or import the all module to include everything.

Available Modules:

• common: Basic utility module used by all algorithm modules.
• bom: Dependency management module.
• face: Face recognition and processing features.
• vision: General vision features (e.g., object detection).
• ocr: Optical Character Recognition features.
• translate: Machine translation features.
• speech: Speech features (includes ASR and TTS).

Usage Pattern: To minimize dependency footprint, import only the specific modules you need. To get the full suite, use the all module.

SmartJavaAI is designed to bridge the gap between the Python-centric AI ecosystem (PyTorch/TensorFlow) and Java engineering environments. It addresses two main pain points:

1. Ecosystem Fragmentation: Resolves the disconnect between mainstream deep learning frameworks (Python-based) and Java business logic.
2. Integration Complexity: Replaces scattered, inconsistently encapsulated AI solutions with a unified, low-barrier interface.

Key Features:

• Out-of-the-box: Call complex algorithms with as little as two lines of code.
• Multi-Engine Support: Compatible with PyTorch, TensorFlow, MXNet, and ONNX Runtime.
• Cross-Platform: Supports Windows, Linux, and macOS on both x86 and ARM architectures.