SmartJavaAI: A Java AI Toolkit Built on DJL

[geekwenjie]

Hi DJL community 👋,

I’d like to share a project I’ve been working on: SmartJavaAI,
a lightweight and easy-to-use Java AI toolkit built on top of DJL.

🔹 What is SmartJavaAI?

SmartJavaAI aims to make AI integration in Java projects easier. It wraps multiple models under a unified API so that Java developers can use state-of-the-art AI features without diving deep into Python code or model internals.

🔹 Modules currently available

Face Recognition

• Face Detection: 5-point facial landmark detection
• Face Recognition: 512-dimensional feature extraction, face alignment, 1:1 face comparison, 1:N face identification
• Face Database: Face registration, update, query, deletion (supports vector databases like Milvus/SQLite)
• Face Attribute Detection: Gender, age, mask, eye state, facial pose
• Silent Liveness Detection: Image and video liveness detection
• Facial Expression Recognition: 7 types of expressions
• Face Quality Assessment: Brightness, sharpness, completeness, pose, and occlusion evaluation

Object Detection

• Supports multiple mainstream models: Compatible with YOLOv3, YOLOv5, YOLOv8, YOLOv11, YOLOv12, SSD, and other object detection algorithms
• Custom model loading: Seamlessly load and deploy user-trained object detection models

OCR (Text Recognition)

• Supports PaddleOCR 3.0 models: Integrates the latest PP-OCRv5, PP-OCRv4, table structure recognition models (SLANet_plus), and text line orientation classification models
• Arbitrary angle recognition & orientation correction
• General text recognition and handwriting recognition
• Table recognition
• Chinese license plate recognition: Single-layer/double-layer detection, color recognition, supports 12 types of Chinese license plates

Machine Translation

• NLLB-200 model integration: Supports translation among 200+ languages

Speech Recognition

• OpenAI Whisper model integration: Supports 100 languages
• Vosk speech recognition integration

---

🔹 Why SmartJavaAI?

• Includes many model examples that are not yet officially covered by DJL (e.g., some face recognition and OCR pipelines).
• Provides simple and elegant Java APIs that wrap model loading, inference, and post-processing.
• Designed as a general JAR library that can be integrated easily into Spring Boot or any Java project.

---

I believe this can be helpful for Java developers who want to leverage DJL with less effort.
The project is open source, and contributions, feedback, and discussions are more than welcome! 🚀

👉 GitHub: [https://github.com/geekwenjie/SmartJavaAI)