临时提交

README.md

@@ -228,7 +228,8 @@ SmartJavaAI是专为JAVA 开发者打造的一个功能丰富、开箱即用的
       <td>
         <div align="left">
           <p>语音识别</p>
-          - 支持100种语言
+          - 支持100种语言<br>
+          - 支持实时语音识别
         </div>
       </td>     
       <td>
@@ -343,7 +344,7 @@ SmartJavaAI是专为JAVA 开发者打造的一个功能丰富、开箱即用的
 <dependency>
     <groupId>cn.smartjavaai</groupId>
     <artifactId>smartjavaai-all</artifactId>
-    <version>1.0.23</version>
+    <version>1.0.24</version>
 </dependency>
 ```
 ### 3、完整示例代码
@@ -592,15 +593,20 @@ SmartJavaAI是专为JAVA 开发者打造的一个功能丰富、开箱即用的
 ## 献代码的步骤
 
 1、在Gitee或者Github/Gitcode上fork项目到自己的repo
+
 2、把fork过去的项目也就是你的项目clone到你的本地
+
 3、修改代码（记得一定要修改dev分支）
+
 4、commit后push到自己的库（dev分支）
+
 5、登录Gitee或Github/Gitcode在你首页可以看到一个 pull request 按钮，点击它，填写一些说明信息，然后提交即可。
+
 6、等待维护者合并
 
 ## 近期更新日志
 
-## [v1.0.23] - 2025-08-09
+## [v1.0.24] - 2025-08-09
 - 新增 语音识别模块，集成 OpenAI 开源的 Whisper 和 Vosk
 - 修复 质量评估模型的 Bug
 - 修复 OCR 模块 recognizeAndDraw 方法的 Bug

all/pom.xml

@@ -6,11 +6,11 @@
     <parent>
         <groupId>cn.smartjavaai</groupId>
         <artifactId>smartjavaai-parent</artifactId>
-        <version>1.0.23</version>
+        <version>1.0.24</version>
     </parent>
 
-    <artifactId>smartjavaai-all</artifactId>
-    <version>1.0.23</version>
+    <artifactId>all</artifactId>
+    <version>1.0.24</version>
     <name>${project.artifactId}</name>
     <description>SmartJavaAI</description>
     <url>https://github.com/geekwenjie/SmartJavaAI</url>
@@ -33,31 +33,31 @@
     <dependencies>
         <dependency>
             <groupId>cn.smartjavaai</groupId>
-            <artifactId>smartjavaai-face</artifactId>
+            <artifactId>face</artifactId>
             <version>${project.version}</version>
         </dependency>
 
         <dependency>
             <groupId>cn.smartjavaai</groupId>
-            <artifactId>smartjavaai-objectdetection</artifactId>
+            <artifactId>vision</artifactId>
             <version>${project.version}</version>
         </dependency>
 
         <dependency>
             <groupId>cn.smartjavaai</groupId>
-            <artifactId>smartjavaai-ocr</artifactId>
+            <artifactId>ocr</artifactId>
             <version>${project.version}</version>
         </dependency>
 
         <dependency>
             <groupId>cn.smartjavaai</groupId>
-            <artifactId>smartjavaai-translate</artifactId>
+            <artifactId>translate</artifactId>
             <version>${project.version}</version>
         </dependency>
 
         <dependency>
             <groupId>cn.smartjavaai</groupId>
-            <artifactId>smartjavaai-speech</artifactId>
+            <artifactId>speech</artifactId>
             <version>${project.version}</version>
         </dependency>
 

all/src/test/java/Test.java

@@ -0,0 +1,47 @@
+import ai.djl.Application;
+import ai.djl.Model;
+import ai.djl.repository.zoo.Criteria;
+import ai.djl.repository.zoo.ModelZoo;
+import cn.smartjavaai.ocr.config.OcrDetModelConfig;
+import cn.smartjavaai.ocr.entity.OcrBox;
+import cn.smartjavaai.ocr.enums.CommonDetModelEnum;
+import cn.smartjavaai.ocr.factory.OcrModelFactory;
+import cn.smartjavaai.ocr.model.common.detect.OcrCommonDetModel;
+import lombok.extern.slf4j.Slf4j;
+import org.bytedeco.javacv.FFmpegFrameGrabber;
+import org.bytedeco.javacv.Frame;
+import org.bytedeco.javacv.Java2DFrameUtils;
+
+import java.awt.image.BufferedImage;
+import java.io.IOException;
+import java.util.List;
+
+/**
+ * @author dwj
+ * @date 2025/4/24
+ */
+@Slf4j
+public class Test {
+
+
+    public static String savePath = "/Users/wenjie/Downloads/";
+    //public static String image1Path = "/Users/wenjie/Documents/idea_workplace/SmartJavaAI-Demo/src/main/resources/5.jpg";
+    public static String image1Path = "/Users/wenjie/Documents/idea_workplace/SmartJavaAI-Demo/src/main/resources/MJ_20250226_172200.png";
+
+    public static String image2Path = "/Users/wenjie/Documents/idea_workplace/SmartJavaAI-Demo/src/main/resources/MJ_20250226_172222.png";
+
+    public static void main(String[] args) throws IOException {
+//        // 加载模型
+//        Model model = ModelZoo.loadModel(Criteria.builder()
+//                .optApplication(Application.NLP.ANY)
+//                .optEngine("PyTorch")
+//                .optModelName("Llama 3")
+//                        .optTranslatorFactory(new Llama3TranslatorFactory())
+//                .optTranslatorProvider(() -> new Llama3Translator())
+//                .build());
+
+
+    }
+
+
+}

bom/pom.xml

@@ -6,12 +6,12 @@
     <parent>
         <groupId>cn.smartjavaai</groupId>
         <artifactId>smartjavaai-parent</artifactId>
-        <version>1.0.23</version>
+        <version>1.0.24</version>
     </parent>
 
-    <version>1.0.23</version>
-    <artifactId>smartjavaai-bom</artifactId>
-    <name>smartjavaai-bom</name>
+    <version>1.0.24</version>
+    <artifactId>bom</artifactId>
+    <name>sbom</name>
     <description>统一版本管理的 BOM 包，同时支持 import 和全量依赖</description>
 
     <properties>
@@ -25,27 +25,27 @@
         <dependencies>
             <dependency>
                 <groupId>cn.smartjavaai</groupId>
-                <artifactId>smartjavaai-face</artifactId>
+                <artifactId>face</artifactId>
                 <version>${project.parent.version}</version>
             </dependency>
             <dependency>
                 <groupId>cn.smartjavaai</groupId>
-                <artifactId>smartjavaai-objectdetection</artifactId>
+                <artifactId>vision</artifactId>
                 <version>${project.parent.version}</version>
             </dependency>
             <dependency>
                 <groupId>cn.smartjavaai</groupId>
-                <artifactId>smartjavaai-ocr</artifactId>
+                <artifactId>ocr</artifactId>
                 <version>${project.parent.version}</version>
             </dependency>
             <dependency>
                 <groupId>cn.smartjavaai</groupId>
-                <artifactId>smartjavaai-translate</artifactId>
+                <artifactId>translate</artifactId>
                 <version>${project.parent.version}</version>
             </dependency>
             <dependency>
                 <groupId>cn.smartjavaai</groupId>
-                <artifactId>smartjavaai-speech</artifactId>
+                <artifactId>speech</artifactId>
                 <version>${project.parent.version}</version>
             </dependency>
         </dependencies>

common/pom.xml

@@ -6,11 +6,11 @@
     <parent>
         <groupId>cn.smartjavaai</groupId>
         <artifactId>smartjavaai-parent</artifactId>
-        <version>1.0.23</version>
+        <version>1.0.24</version>
     </parent>
 
-    <name>smartjavaai-common</name>
-    <artifactId>smartjavaai-common</artifactId>
+    <name>common</name>
+    <artifactId>common</artifactId>
     <description>SmartJavaAI</description>
     <url>https://github.com/geekwenjie/SmartJavaAI</url>
     <licenses>

common/src/main/java/cn/smartjavaai/common/config/Config.java

@@ -8,6 +8,7 @@ import lombok.extern.slf4j.Slf4j;
 import org.apache.commons.lang3.StringUtils;
 
 import java.io.File;
+import java.nio.file.Paths;
 
 /**
  * 全局配置
@@ -68,16 +69,25 @@ public class Config {
         String osName = SystemUtil.getOsInfo().getName();
         log.info("当前操作系统：{}", osName);
         if(osName.toLowerCase().contains("windows")){
-            cachePath = SystemUtil.getUserInfo().getHomeDir() + CACHE_DIR;
+            cachePath = Paths.get(
+                    SystemUtil.getUserInfo().getHomeDir(),
+                    "smartjavaai_cache"
+            ).toString();
             FileUtil.mkdir(cachePath);
         }else if(osName.toLowerCase().contains("linux")){
             cachePath = "/root/" + CACHE_DIR;
             FileUtil.mkdir(cachePath);
         }else if(osName.toLowerCase().contains("mac")){
-            cachePath = SystemUtil.getUserInfo().getHomeDir() + CACHE_DIR;
+            cachePath = Paths.get(
+                    SystemUtil.getUserInfo().getHomeDir(),
+                    "smartjavaai_cache"
+            ).toString();
             FileUtil.mkdir(cachePath);
         }else{
-            cachePath = SystemUtil.getUserInfo().getHomeDir() + CACHE_DIR;
+            cachePath = Paths.get(
+                    SystemUtil.getUserInfo().getHomeDir(),
+                    "smartjavaai_cache"
+            ).toString();
             FileUtil.mkdir(cachePath);
         }
     }

common/src/main/java/cn/smartjavaai/common/cv/SmartImageFactory.java

@@ -0,0 +1,65 @@
+package cn.smartjavaai.common.cv;
+
+import ai.djl.modality.cv.BufferedImageFactory;
+import ai.djl.modality.cv.Image;
+import ai.djl.modality.cv.ImageFactory;
+import ai.djl.modality.cv.output.DetectedObjects;
+import ai.djl.modality.cv.util.NDImageUtils;
+import ai.djl.ndarray.NDArray;
+import ai.djl.ndarray.types.DataType;
+import ai.djl.ndarray.types.Shape;
+import ai.djl.opencv.OpenCVImageFactory;
+import ai.djl.util.Utils;
+import cn.smartjavaai.common.utils.Base64ImageUtils;
+import cn.smartjavaai.common.utils.OpenCVUtils;
+import nu.pattern.OpenCV;
+import org.opencv.core.CvType;
+import org.opencv.core.Mat;
+import org.opencv.core.MatOfByte;
+import org.opencv.imgcodecs.Imgcodecs;
+import org.opencv.imgproc.Imgproc;
+
+import java.awt.image.BufferedImage;
+import java.io.ByteArrayInputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.ByteBuffer;
+import java.nio.IntBuffer;
+import java.nio.file.Path;
+
+/**
+ * 图片处理工厂类
+ * @author dwj
+ */
+public class SmartImageFactory extends BufferedImageFactory {
+
+    private static volatile SmartImageFactory instance;
+
+    public static SmartImageFactory newInstance() {
+        if (instance == null) {
+            synchronized (SmartImageFactory.class) {
+                if (instance == null) {
+                    instance = new SmartImageFactory();
+                }
+            }
+        }
+        return instance;
+    }
+
+    public static SmartImageFactory getInstance(){
+        return newInstance();
+    }
+
+    public Image fromBufferedImage(BufferedImage sourceImage){
+        return fromImage(OpenCVUtils.image2Mat(sourceImage));
+    }
+
+    public Image fromBase64(String base64Image) throws IOException {
+        return fromUrl(base64Image);
+    }
+
+    public Image fromBytes(byte[] imageData){
+        return fromImage(new ByteArrayInputStream(imageData));
+    }
+
+}

common/src/main/java/cn/smartjavaai/common/entity/DetectionInfo.java

@@ -32,6 +32,16 @@ public class DetectionInfo {
      */
     private ObjectDetInfo objectDetInfo;
 
+    /**
+     * 目标分割信息
+     */
+    private InstanceSegInfo instanceSegInfo;
+
+    /**
+     * 旋转框信息
+     */
+    private ObbDetInfo obbDetInfo;
+
 
 
     public DetectionInfo() {

common/src/main/java/cn/smartjavaai/common/entity/DetectionResponse.java

@@ -1,5 +1,6 @@
 package cn.smartjavaai.common.entity;
 
+import ai.djl.modality.cv.Image;
 import lombok.Data;
 
 import java.util.List;
@@ -14,6 +15,8 @@ public class DetectionResponse {
 
     private List<DetectionInfo> detectionInfoList;
 
+    private Image drawnImage;
+
 
     public DetectionResponse() {
     }

common/src/main/java/cn/smartjavaai/common/entity/InstanceSegInfo.java

@@ -0,0 +1,29 @@
+package cn.smartjavaai.common.entity;
+
+import lombok.Data;
+
+/**
+ * 目标分割信息
+ * @author dwj
+ */
+@Data
+public class InstanceSegInfo {
+
+    /**
+     * 类别名称
+     */
+    private String className;
+
+    /**
+     * 遮罩
+     */
+    private float[][] mask;
+
+    public InstanceSegInfo() {
+    }
+
+    public InstanceSegInfo(String className, float[][] mask) {
+        this.className = className;
+        this.mask = mask;
+    }
+}

common/src/main/java/cn/smartjavaai/common/entity/ObbDetInfo.java

@@ -0,0 +1,29 @@
+package cn.smartjavaai.common.entity;
+
+import java.util.List;
+
+/**
+ * 定向边界框 检测结果
+ * @author dwj
+ */
+public class ObbDetInfo {
+
+    /**
+     * 类别名称
+     */
+    private String className;
+
+    /**
+     * 检测框坐标
+     */
+    private RotatedBox rotatedBox;
+
+    public ObbDetInfo() {
+    }
+
+
+    public ObbDetInfo(String className, RotatedBox rotatedBox) {
+        this.className = className;
+        this.rotatedBox = rotatedBox;
+    }
+}

common/src/main/java/cn/smartjavaai/common/entity/RotatedBox.java

@@ -0,0 +1,38 @@
+package cn.smartjavaai.common.entity;
+
+/**
+ * 旋转框
+ * @author dwj
+ */
+public class RotatedBox {
+
+    /**
+     * 左上角
+     */
+    private Point topLeft;
+
+    /**
+     * 右上角
+     */
+    private Point topRight;
+
+    /**
+     * 右下角
+     */
+    private Point bottomRight;
+
+    /**
+     * 左下角
+     */
+    private Point bottomLeft;
+
+    public RotatedBox(Point topLeft, Point topRight, Point bottomRight, Point bottomLeft) {
+        this.topLeft = topLeft;
+        this.topRight = topRight;
+        this.bottomRight = bottomRight;
+        this.bottomLeft = bottomLeft;
+    }
+
+    public RotatedBox() {
+    }
+}

common/src/main/java/cn/smartjavaai/common/enums/VideoSourceType.java

@@ -0,0 +1,12 @@
+package cn.smartjavaai.common.enums;
+
+/**
+ * 视频源类型枚举
+ * @author dwj
+ * @date 2025/8/27
+ */
+public enum VideoSourceType {
+    STREAM,     // RTSP 或 HTTP 流
+    FILE,       // 本地视频文件
+    CAMERA;      // 本地摄像头
+}

common/src/main/java/cn/smartjavaai/common/utils/DJLCommonUtils.java

@@ -0,0 +1,32 @@
+package cn.smartjavaai.common.utils;
+
+import java.nio.file.Files;
+import java.nio.file.Path;
+
+/**
+ * @author dwj
+ */
+public class DJLCommonUtils {
+
+    /**
+     * 检查模型目录中是否存在 "serving.properties" 文件
+     *
+     * @param modelPath 模型目录路径
+     * @return true 表示存在，false 表示不存在
+     */
+    public static boolean isServingPropertiesExists(Path modelPath) {
+        if (modelPath == null || !Files.exists(modelPath)) {
+            return false;
+        }
+        // 确定目录路径
+        Path dirPath = Files.isDirectory(modelPath) ? modelPath : modelPath.getParent();
+        if (dirPath == null) {
+            return false; // 可能是根目录的文件
+        }
+
+        // 判断目录下的 serving.properties 是否存在
+        Path servingFile = dirPath.resolve("serving.properties");
+        return Files.exists(servingFile);
+    }
+
+}

common/src/main/java/cn/smartjavaai/common/utils/FrameConverterUtil.java

@@ -0,0 +1,66 @@
+package cn.smartjavaai.common.utils;
+
+import ai.djl.modality.cv.Image;
+import ai.djl.modality.cv.ImageFactory;
+import org.bytedeco.javacpp.BytePointer;
+import org.bytedeco.javacv.OpenCVFrameConverter;
+import org.bytedeco.opencv.opencv_core.CvMat;
+import org.bytedeco.opencv.opencv_core.Mat;
+import org.opencv.core.CvType;
+
+import java.awt.image.BufferedImage;
+
+/**
+ * @author dwj
+ * @date 2025/8/27
+ */
+public class FrameConverterUtil {
+
+    /**
+     * 将 Bytedeco Mat 转为 DJL Image
+     * 支持 1/3/4 通道
+     */
+    public static Image matToDJLImage(Mat cvMat) {
+        if (cvMat == null || cvMat.empty()) {
+            return null;
+        }
+
+        int width = cvMat.cols();
+        int height = cvMat.rows();
+        int channels = cvMat.channels();
+
+        int[] pixels = new int[width * height];
+
+        if (channels == 1) { // 灰度图
+            byte[] data = new byte[width * height];
+            cvMat.data().get(data);
+            for (int i = 0; i < width * height; i++) {
+                int gray = data[i] & 0xFF;
+                pixels[i] = (0xFF << 24) | (gray << 16) | (gray << 8) | gray;
+            }
+        } else if (channels == 3) { // BGR
+            byte[] data = new byte[width * height * 3];
+            cvMat.data().get(data);
+            for (int i = 0; i < width * height; i++) {
+                int b = data[i * 3] & 0xFF;
+                int g = data[i * 3 + 1] & 0xFF;
+                int r = data[i * 3 + 2] & 0xFF;
+                pixels[i] = (0xFF << 24) | (r << 16) | (g << 8) | b;
+            }
+        } else if (channels == 4) { // BGRA
+            byte[] data = new byte[width * height * 4];
+            cvMat.data().get(data);
+            for (int i = 0; i < width * height; i++) {
+                int b = data[i * 4] & 0xFF;
+                int g = data[i * 4 + 1] & 0xFF;
+                int r = data[i * 4 + 2] & 0xFF;
+                int a = data[i * 4 + 3] & 0xFF;
+                pixels[i] = (a << 24) | (r << 16) | (g << 8) | b;
+            }
+        } else {
+            throw new IllegalArgumentException("只支持 1/3/4 通道图像");
+        }
+
+        return ImageFactory.getInstance().fromPixels(pixels, width, height);
+    }
+}

common/src/main/java/cn/smartjavaai/common/utils/ImageUtils.java

@@ -139,7 +139,7 @@ public class ImageUtils {
      * @param width
      * @param height
      */
-    public static void drawImageRect(Image image, int x, int y, int width, int height) {
+    public static void drawBufferedImageRect(Image image, int x, int y, int width, int height) {
         // 将绘制图像转换为Graphics2D
         BufferedImage bufferedImage = (BufferedImage)image.getWrappedImage();
         Graphics2D g = (Graphics2D) bufferedImage.getGraphics();
@@ -501,4 +501,7 @@ public class ImageUtils {
 
 
 
+
+
+
 }

common/src/main/java/cn/smartjavaai/common/utils/LetterBoxUtils.java

@@ -1,5 +1,6 @@
 package cn.smartjavaai.common.utils;
 
+import ai.djl.modality.cv.output.Rectangle;
 import ai.djl.modality.cv.util.NDImageUtils;
 import ai.djl.ndarray.NDArray;
 import ai.djl.ndarray.NDManager;
@@ -121,4 +122,27 @@ public class LetterBoxUtils {
         return boxes;
     }
 
+    /**
+     * 恢复缩放后的 box(左上角坐标)
+     * @param rectangle
+     * @param scale
+     * @param origImageWidth
+     * @param origImageHeight
+     */
+    public static Rectangle restoreBox(Rectangle rectangle, float scale, int origImageWidth, int origImageHeight, int inputWidth, int inputHeight){
+        double paddingWidth = (inputWidth - origImageWidth * scale) / 2;
+        double paddingHeight = (inputHeight - origImageHeight * scale) / 2;
+
+        // 去掉 padding
+        double x_noPad = rectangle.getX() - paddingWidth;
+        double y_noPad = rectangle.getY() - paddingHeight;
+
+        //模型输出就是原图坐标
+        double x1 = x_noPad / scale / origImageWidth;
+        double y1 = y_noPad / scale / origImageHeight;
+        double boxW = rectangle.getWidth() / scale / origImageWidth ;
+        double boxH = rectangle.getHeight() / scale / origImageHeight;
+        return new Rectangle(x1, y1, boxW, boxH);
+    }
+
 }

common/src/main/java/cn/smartjavaai/common/utils/NMSUtils.java

@@ -1,6 +1,10 @@
 package cn.smartjavaai.common.utils;
 
 import ai.djl.ndarray.NDArray;
+import ai.djl.ndarray.NDArrays;
+import ai.djl.ndarray.NDList;
+import ai.djl.ndarray.NDManager;
+import ai.djl.ndarray.types.DataType;
 
 import java.util.ArrayList;
 import java.util.List;
@@ -64,4 +68,43 @@ public class NMSUtils {
         return keep.stream().mapToInt(i -> i).toArray();
     }
 
+
+
+    /**
+     * 批量执行 NMS，输入 NDArray 形式的 boxes、scores 和 idxs，返回保留的索引列表
+     *
+     * @param boxes   NDArray 形状为 (N, 4)，格式为 [x1, y1, x2, y2]
+     * @param scores  NDArray 形状为 (N,) 或 (N,1)，每个 box 的置信度
+     * @param idxs    NDArray 形状为 (N,)，每个 box 对应的 batch id
+     * @param iouThreshold IOU 阈值，超过该阈值则认为有
+     * @return 批量保留框的索引列表
+     *
+     */
+    public static NDArray batchedNms(NDArray boxes, NDArray scores, NDArray idxs, float iouThreshold, NDManager manager) {
+        List<NDArray> keepList = new ArrayList<>();
+        // 获取唯一 batch id
+        NDArray uniqueIdxs = idxs.unique().get(0);
+        for (long batchId : uniqueIdxs.toLongArray()) {
+            // 找出当前 batch 的框
+            NDArray mask = idxs.eq(batchId);
+            NDArray batchBoxes = boxes.get(mask);
+            NDArray batchScores = scores.get(mask);
+            // 执行单 batch NMS
+            int[] keepIndices = nms(batchBoxes, batchScores, iouThreshold);
+            if (keepIndices.length > 0) {
+                // 将局部索引映射回全局索引
+                NDArray globalIndices = manager.arange(boxes.getShape().get(0))
+                        .get(mask)
+                        .toType(DataType.INT64, false)
+                        .get(manager.create(keepIndices));
+
+                keepList.add(globalIndices);
+            }
+        }
+        if (keepList.isEmpty()) {
+            return manager.create(new long[0]);
+        }
+        return NDArrays.concat(new NDList(keepList));
+    }
+
 }

common/src/main/java/cn/smartjavaai/common/utils/OpenCVUtils.java

@@ -0,0 +1,265 @@
+package cn.smartjavaai.common.utils;
+
+import ai.djl.modality.cv.Image;
+import ai.djl.modality.cv.ImageFactory;
+import ai.djl.ndarray.NDArray;
+import ai.djl.ndarray.NDManager;
+import ai.djl.util.RandomUtils;
+import cn.smartjavaai.common.entity.DetectionInfo;
+import org.apache.commons.collections.CollectionUtils;
+import org.apache.commons.lang3.StringUtils;
+import org.opencv.core.*;
+import org.opencv.core.Point;
+import org.opencv.imgproc.Imgproc;
+
+import java.awt.*;
+import java.awt.image.BufferedImage;
+import java.awt.image.DataBufferByte;
+import java.util.List;
+import java.util.Objects;
+
+/**
+ * OpenCV 工具类
+ */
+public class OpenCVUtils {
+    /**
+     * canny算法，边缘检测
+     *
+     * @param src
+     * @return
+     */
+    public static Mat canny(Mat src) {
+        Mat mat = src.clone();
+        Imgproc.Canny(src, mat, 100, 200);
+        return mat;
+    }
+
+    /**
+     * 画线
+     *
+     * @param mat
+     * @param point1
+     * @param point2
+     */
+    public static void line(Mat mat, Point point1, Point point2) {
+        Imgproc.line(mat, point1, point2, new Scalar(255, 255, 255), 1);
+    }
+
+    /**
+     * NDArray to opencv_core.Mat
+     *
+     * @param manager
+     * @param srcPoints
+     * @param dstPoints
+     * @return
+     */
+    public static Mat toOpenCVMat(NDManager manager, NDArray srcPoints, NDArray dstPoints) {
+        NDArray svdMat = SVDUtils.transformationFromPoints(manager, srcPoints, dstPoints);
+        double[] doubleArray = svdMat.toDoubleArray();
+        Mat newSvdMat = new Mat(2, 3, CvType.CV_64F);
+        for (int i = 0; i < 2; i++) {
+            for (int j = 0; j < 3; j++) {
+                newSvdMat.put(i, j, doubleArray[i * 3 + j]);
+            }
+        }
+        return newSvdMat;
+    }
+
+    /**
+     * double[][] points array to Mat
+     * @param points
+     * @return
+     */
+    public static Mat toOpenCVMat(double[][] points) {
+        Mat mat = new Mat(5, 2, CvType.CV_64F);
+        for (int i = 0; i < 5; i++) {
+            for (int j = 0; j < 2; j++) {
+                mat.put(i, j, points[i * 5 + j]);
+            }
+        }
+        return mat;
+    }
+
+    /**
+     * 变换矩阵的逆矩阵
+     *
+     * @param src
+     * @return
+     */
+    public static Mat invertAffineTransform(Mat src) {
+        Mat dst = src.clone();
+        Imgproc.invertAffineTransform(src, dst);
+        return dst;
+    }
+
+    /**
+     * Mat to BufferedImage
+     *
+     * @param mat
+     * @return
+     */
+    public static BufferedImage mat2Image(Mat mat) {
+        int width = mat.width();
+        int height = mat.height();
+        byte[] data = new byte[width * height * (int) mat.elemSize()];
+        Imgproc.cvtColor(mat, mat, 4);
+        mat.get(0, 0, data);
+        BufferedImage ret = new BufferedImage(width, height, 5);
+        ret.getRaster().setDataElements(0, 0, width, height, data);
+        return ret;
+    }
+
+    /**
+     * BufferedImage to Mat
+     *
+     * @param img
+     * @return
+     */
+    public static Mat image2Mat(BufferedImage img) {
+        int width = img.getWidth();
+        int height = img.getHeight();
+
+        // 强制转换为 TYPE_3BYTE_BGR，自动去除透明通道
+        BufferedImage convertedImg = new BufferedImage(width, height, BufferedImage.TYPE_3BYTE_BGR);
+        Graphics2D g2d = convertedImg.createGraphics();
+        g2d.drawImage(img, 0, 0, null);
+        g2d.dispose();
+
+        byte[] data = ((DataBufferByte) convertedImg.getRaster().getDataBuffer()).getData();
+        Mat mat = new Mat(height, width, CvType.CV_8UC3);
+        mat.put(0, 0, data);
+        return mat;
+    }
+
+    /**
+     * 透视变换
+     *
+     * @param src
+     * @param srcPoints
+     * @param dstPoints
+     * @return
+     */
+    public static Mat perspectiveTransform(Mat src, Mat srcPoints, Mat dstPoints) {
+        Mat dst = src.clone();
+        Mat warp_mat = Imgproc.getPerspectiveTransform(srcPoints, dstPoints);
+        Imgproc.warpPerspective(src, dst, warp_mat, dst.size());
+        warp_mat.release();
+        return dst;
+    }
+
+    /**
+     * 绘制矩形框和文字
+     *
+     * @param image
+     * @param detectionInfoList
+     */
+    public static void drawRectAndText(Image image, List<DetectionInfo> detectionInfoList) {
+        if(CollectionUtils.isEmpty(detectionInfoList))
+            return;
+        for(DetectionInfo detectionInfo : detectionInfoList){
+            drawRectAndText(image, detectionInfo);
+        }
+    }
+
+
+    /**
+     * 绘制矩形框和文字
+     *
+     * @param image
+     * @param detectionInfo
+     */
+    public static void drawRectAndText(Image image, DetectionInfo detectionInfo) {
+
+
+        Mat mat = (Mat)image.getWrappedImage();
+        if (image == null) return;
+        int x = detectionInfo.getDetectionRectangle().getX();
+        int y = detectionInfo.getDetectionRectangle().getY();
+        int width = detectionInfo.getDetectionRectangle().getWidth();
+        int height = detectionInfo.getDetectionRectangle().getHeight();
+
+        Scalar rectangleColor = new Scalar((double)RandomUtils.nextInt(178), (double)RandomUtils.nextInt(178), (double)RandomUtils.nextInt(178));
+
+        // 绘制矩形框
+        Point pt1 = new Point(x, y);
+        Point pt2 = new Point(x + width, y + height);
+        Imgproc.rectangle(mat, pt1, pt2, rectangleColor, 2);
+
+        // 绘制文字
+        if (Objects.nonNull(detectionInfo.getObjectDetInfo()) && StringUtils.isNotBlank(detectionInfo.getObjectDetInfo().getClassName())) {
+            String className = detectionInfo.getObjectDetInfo().getClassName();
+            Size size = Imgproc.getTextSize(className, 1, 1.3, 1, (int[])null);
+            Point br = new Point((double)x + size.width + 4.0, (double)y + size.height + 4.0);
+            Imgproc.rectangle(mat, pt1, br, rectangleColor, -1);
+            Point point = new Point((double)x, (double)y + size.height + 2.0);
+            Scalar color = new Scalar(255.0, 255.0, 255.0);
+            Imgproc.putText(mat, className, point, 1, 1.3, color, 1);
+        }
+        image = ImageFactory.getInstance().fromImage(mat);
+    }
+
+    /**
+     * 在Mat上绘制矩形框和文字
+     *
+     * @param mat       待绘制的Mat
+     * @param x         矩形左上角X
+     * @param y         矩形左上角Y
+     * @param width     矩形宽度
+     * @param height    矩形高度
+     * @param color     框的颜色，例如 new Scalar(0, 255, 0) 绿色
+     * @param thickness 框线宽度
+     * @param text      需要绘制的文字，可以为null或空
+     * @param fontScale 文字缩放比例
+     * @param textColor 文字颜色
+     */
+    public static void drawRectAndText(Mat mat,
+                                       int x, int y, int width, int height,
+                                       Scalar color, int thickness,
+                                       String text, double fontScale, Scalar textColor) {
+
+        if (mat == null || mat.empty()) return;
+
+        // 绘制矩形框
+        Point pt1 = new Point(x, y);
+        Point pt2 = new Point(x + width, y + height);
+        Imgproc.rectangle(mat, pt1, pt2, color, thickness);
+
+        // 绘制文字
+        if (text != null && !text.isEmpty()) {
+            int baseline[] = new int[1];
+            Size textSize = Imgproc.getTextSize(text, Imgproc.FONT_HERSHEY_SIMPLEX, fontScale, thickness, baseline);
+            // 保证文字不超出矩形
+            Point textOrg = new Point(x, y - 5 < 0 ? y + textSize.height + 5 : y - 5);
+            Imgproc.putText(mat, text, textOrg, Imgproc.FONT_HERSHEY_SIMPLEX, fontScale, textColor, thickness);
+        }
+    }
+
+    /**
+     * 将 Bytedeco 的 Mat 转换为 OpenCV 官方的 Mat
+     * @param src Bytedeco Mat (BGR 或 BGRA)
+     * @return OpenCV Mat (BGR 或 BGRA)
+     */
+    public static org.opencv.core.Mat convertToOpenCVMat(org.bytedeco.opencv.opencv_core.Mat bMat) {
+
+
+        try {
+            int width = bMat.cols();
+            int height = bMat.rows();
+            int channels = bMat.channels();
+
+            // 创建 OpenCV Mat
+            org.opencv.core.Mat cvMat = new org.opencv.core.Mat(height, width, channels == 3 ? CvType.CV_8UC3 : CvType.CV_8UC1);
+
+            // 从 bytedeco Mat 获取像素数据
+            byte[] data = new byte[width * height * channels];
+            bMat.data().get(data);
+
+            // 填充到 OpenCV Mat
+            cvMat.put(0, 0, data);
+            return cvMat;
+        } catch (Throwable e) {
+            e.printStackTrace();
+        }
+        return null;
+    }
+}

examples/face-example/pom.xml

@@ -12,7 +12,7 @@
         <maven.compiler.source>11</maven.compiler.source>
         <maven.compiler.target>11</maven.compiler.target>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
-        <smartjavaai.version>1.0.23</smartjavaai.version>
+        <smartjavaai.version>1.0.24</smartjavaai.version>
         <!--如果打包运行，需要替换成你的main-->
         <exec.mainClass>smartai.examples.face.facedet.FaceDetDemo</exec.mainClass>
 
@@ -255,6 +255,14 @@
             <scope>runtime</scope>
         </dependency>
 
+        <dependency>
+            <groupId>ai.djl.pytorch</groupId>
+            <artifactId>pytorch-native-cpu</artifactId>
+            <classifier>linux-aarch64</classifier>
+            <scope>runtime</scope>
+            <version>2.5.1</version>
+        </dependency>
+
 
 
     </dependencies>

examples/face-example/src/main/java/smartai/examples/face/PythonTranslator.java

@@ -0,0 +1,118 @@
+/*
+ * Copyright 2023 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License"). You may not use this file except in compliance
+ * with the License. A copy of the License is located at
+ *
+ * http://aws.amazon.com/apache2.0/
+ *
+ * or in the "license" file accompanying this file. This file is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES
+ * OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions
+ * and limitations under the License.
+ */
+package smartai.examples.face;
+
+import ai.djl.ModelException;
+import ai.djl.inference.Predictor;
+import ai.djl.modality.Classifications;
+import ai.djl.modality.Input;
+import ai.djl.modality.Output;
+import ai.djl.modality.cv.Image;
+import ai.djl.ndarray.NDList;
+import ai.djl.repository.zoo.Criteria;
+import ai.djl.repository.zoo.ZooModel;
+import ai.djl.translate.NoBatchifyTranslator;
+import ai.djl.translate.TranslateException;
+import ai.djl.translate.TranslatorContext;
+import ai.djl.util.JsonUtils;
+import ai.djl.util.Utils;
+
+import com.google.gson.reflect.TypeToken;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.lang.reflect.Type;
+import java.net.URL;
+import java.nio.file.Paths;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+
+public class PythonTranslator implements NoBatchifyTranslator<byte[], Classifications> {
+
+    private ZooModel<Input, Output> model;
+    private Predictor<Input, Output> predictor;
+
+    @Override
+    public void prepare(TranslatorContext ctx) throws ModelException, IOException {
+        if (predictor == null) {
+            Criteria<Input, Output> criteria =
+                    Criteria.builder()
+                            .setTypes(Input.class, Output.class)
+                            .optModelPath(Paths.get("src/test/python"))
+                            .optEngine("Python")
+                            .build();
+            model = criteria.loadModel();
+            predictor = model.newPredictor();
+        }
+    }
+
+//    @Override
+//    public NDList processInput(TranslatorContext ctx, String url)
+//            throws IOException, TranslateException {
+//        Input input = new Input();
+//        try (InputStream is = new URL(url).openStream()) {
+//            input.add("data", Utils.toByteArray(is));
+//        }
+//        input.addProperty("Content-Type", "image/jpeg");
+//        // calling preprocess() function in model.py
+//        input.addProperty("handler", "preprocess");
+//        Output output = predictor.predict(input);
+//        if (output.getCode() != 200) {
+//            throw new TranslateException("Python preprocess() failed: " + output.getMessage());
+//        }
+//
+//        return output.getDataAsNDList(ctx.getNDManager());
+//    }
+
+    @Override
+    public NDList processInput(TranslatorContext ctx, byte[] image)
+            throws IOException, TranslateException {
+        Input input = new Input();
+        input.add("data", image);
+        input.addProperty("Content-Type", "image/jpeg");
+        // calling preprocess() function in model.py
+        input.addProperty("handler", "preprocess");
+        Output output = predictor.predict(input);
+        if (output.getCode() != 200) {
+            throw new TranslateException("Python preprocess() failed: " + output.getMessage());
+        }
+        return output.getDataAsNDList(ctx.getNDManager());
+    }
+
+    @Override
+    public Classifications processOutput(TranslatorContext ctx, NDList list)
+            throws TranslateException {
+        Input input = new Input();
+        input.add("data", list);
+        // calling postprocess() function in processing.py
+        input.addProperty("handler", "postprocess");
+        Output output = predictor.predict(input);
+        if (output.getCode() != 200) {
+            throw new TranslateException("Python postprocess() failed: " + output.getMessage());
+        }
+
+        String json = output.getData().getAsString();
+        System.out.println("json:" + json);
+        return null;
+    }
+
+    public void close() {
+        if (predictor != null) {
+            predictor.close();
+            model.close();
+            predictor = null;
+            model = null;
+        }
+    }
+}

examples/face-example/src/main/java/smartai/examples/face/Test.java

@@ -0,0 +1,99 @@
+package smartai.examples.face;
+
+import ai.djl.Application;
+import ai.djl.Device;
+import ai.djl.MalformedModelException;
+import ai.djl.inference.Predictor;
+import ai.djl.modality.Classifications;
+import ai.djl.modality.audio.Audio;
+import ai.djl.modality.audio.AudioFactory;
+import ai.djl.modality.audio.translator.SpeechRecognitionTranslatorFactory;
+import ai.djl.repository.Artifact;
+import ai.djl.repository.MRL;
+import ai.djl.repository.zoo.Criteria;
+import ai.djl.repository.zoo.ModelNotFoundException;
+import ai.djl.repository.zoo.ModelZoo;
+import ai.djl.repository.zoo.ZooModel;
+import ai.djl.translate.TranslateException;
+import lombok.extern.slf4j.Slf4j;
+
+import java.io.IOException;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.nio.file.Paths;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * @author dwj
+ * @date 2025/7/29
+ */
+@Slf4j
+public class Test {
+
+    public static void main(String[] args) throws ModelNotFoundException, MalformedModelException, IOException, TranslateException {
+//        PythonTranslator translator = new PythonTranslator();
+//        Criteria<byte[], Classifications> criteria =
+//                Criteria.builder()
+//                        .setTypes(byte[].class, Classifications.class)
+//                        .optModelPath(Paths.get("/Users/wenjie/Documents/develop/model/arcfaceresnet100-11-int8.onnx"))
+//                        .optEngine("OnnxRuntime")
+//                        .optTranslator(translator)
+//                        .build();
+//        String path = "/Users/wenjie/Downloads/facetest/jsy.jpg";
+//        try (ZooModel<byte[], Classifications> model = criteria.loadModel();
+//             Predictor<byte[], Classifications> predictor = model.newPredictor()) {
+//            byte[] data = Files.readAllBytes(Paths.get(path));
+//            Classifications ret = predictor.predict(data);
+//            System.out.println(ret);
+//        }
+//
+//        // unload python model
+//        translator.close();
+
+
+        // Load model.
+        // Wav2Vec2 model is a speech model that accepts a float array corresponding to the raw
+        // waveform of the speech signal.
+
+//        String url = "/Users/wenjie/Downloads/20210601_u2++_conformer_exp/final.pt";
+//        Criteria<Audio, String> criteria =
+//                Criteria.builder()
+//                        .setTypes(Audio.class, String.class)
+////                        .optModelUrls(url)
+//                        .optModelPath(Paths.get(url))
+//                        .optDevice(Device.cpu()) // torchscript model only support CPU
+//                        .optTranslatorFactory(new SpeechRecognitionTranslatorFactory())
+////                        .optModelName("data.pkl")
+//                        .optEngine("PyTorch")
+//                        .build();
+//
+//        // Read in audio file
+//        String wave = "https://resources.djl.ai/audios/speech.wav";
+//        Audio audio = AudioFactory.newInstance().fromUrl(wave);
+//        try (ZooModel<Audio, String> model = criteria.loadModel();
+//             Predictor<Audio, String> predictor = model.newPredictor()) {
+//             String result = predictor.predict(audio);
+//             log.info("Result: {}", result);
+//        }
+
+        boolean withArtifacts =
+                args.length > 0 && ("--artifact".equals(args[0]) || "-a".equals(args[0]));
+        if (!withArtifacts) {
+            log.info("============================================================");
+            log.info("user ./gradlew listModel --args='-a' to show artifact detail");
+            log.info("============================================================");
+        }
+        Map<Application, List<Artifact>> models = ModelZoo.listModels();
+        for (Map.Entry<Application, List<Artifact>> entry : models.entrySet()) {
+            String appName = entry.getKey().toString();
+            for (Artifact artifact : entry.getValue()) {
+                if (withArtifacts) {
+                    log.info("{} djl://{}", appName, artifact);
+                } else {
+                    log.info("{} {}", appName, artifact);
+                }
+            }
+        }
+    }
+}

examples/face-example/src/main/java/smartai/examples/face/facedet/FaceDetDemo.java

@@ -67,7 +67,7 @@ public class FaceDetDemo {
         //高精度模型，速度慢
         config.setModelEnum(FaceDetModelEnum.RETINA_FACE);//人脸检测模型
         //下载模型并替换本地路径，下载地址：https://pan.baidu.com/s/10l22x5fRz_gwLr8EAHa1Jg?pwd=1234 提取码: 1234
-        config.setModelPath("/Users/xxx/Documents/develop/model/retinaface.pt");
+        config.setModelPath("/Users/wenjie/Documents/develop/face_model/retinaface.pt");
         config.setConfidenceThreshold(FaceDetectConstant.DEFAULT_CONFIDENCE_THRESHOLD);//只返回相似度大于该值的人脸
         config.setNmsThresh(FaceDetectConstant.NMS_THRESHOLD);//用于去除重复的人脸框，当两个框的重叠度超过该值时，只保留一个
         return FaceDetModelFactory.getInstance().getModel(config);
@@ -95,12 +95,21 @@ public class FaceDetDemo {
     @Test
     public void testFaceDetect(){
         try {
-            FaceDetModel faceModel = FaceDetModelFactory.getInstance().getModel();
+            FaceDetModel faceModel = getFaceDetModel();
             R<DetectionResponse> detectedResult = faceModel.detect(imgPath);
-            if(detectedResult.isSuccess()){
-                log.info("人脸检测结果：{}", JSONObject.toJSONString(detectedResult.getData()));
+//            if(detectedResult.isSuccess()){
+//                log.info("人脸检测结果：{}", JSONObject.toJSONString(detectedResult.getData()));
+//            }else{
+//                log.info("人脸检测失败：{}", detectedResult.getMessage());
+//            }
+
+            long start = System.currentTimeMillis();
+            R<DetectionResponse> detectedResult2 = faceModel.detect("/Users/wenjie/Downloads/facetest/surprise.png");
+            log.info("耗时：{}", System.currentTimeMillis() - start);
+            if(detectedResult2.isSuccess()){
+                log.info("人脸检测结果：{}", JSONObject.toJSONString(detectedResult2.getData()));
             }else{
-                log.info("人脸检测失败：{}", detectedResult.getMessage());
+                log.info("人脸检测失败：{}", detectedResult2.getMessage());
             }
         } catch (Exception e) {
             e.printStackTrace();

examples/face-example/src/main/java/smartai/examples/face/facerec/FaceRecDemo.java

@@ -63,7 +63,7 @@ public class FaceRecDemo {
         //高精度模型，速度慢
         config.setModelEnum(FaceDetModelEnum.RETINA_FACE);//人脸检测模型
         //下载模型并替换本地路径，下载地址：https://pan.baidu.com/s/10l22x5fRz_gwLr8EAHa1Jg?pwd=1234 提取码: 1234
-        config.setModelPath("/Users/xxx/Documents/develop/model/retinaface.pt");
+//        config.setModelPath("/Users/wenjie/Documents/develop/model/retinaface.pt");
         config.setConfidenceThreshold(FaceDetectConstant.DEFAULT_CONFIDENCE_THRESHOLD);//只返回相似度大于该值的人脸
         config.setNmsThresh(FaceDetectConstant.NMS_THRESHOLD);//用于去除重复的人脸框，当两个框的重叠度超过该值时，只保留一个
         config.setDevice(device);
@@ -170,7 +170,7 @@ public class FaceRecDemo {
         FaceRecConfig config = new FaceRecConfig();
         //高精度模型，速度慢, 追求速度请更换高速模型，具体其他模型参数可以查看文档：http://doc.smartjavaai.cn/face.html
         config.setModelEnum(FaceRecModelEnum.ELASTIC_FACE_MODEL);//人脸检测模型
-        config.setModelPath("/Users/xxx/Documents/develop/model/elasticface.pt");
+        config.setModelPath("/Users/wenjie/Documents/develop/model/elasticface.pt");
         //裁剪人脸：如果图片已经是裁剪过的，则请将此参数设置为false
         config.setCropFace(true);
         //开启人脸对齐：适用于人脸不正的场景，开启将提升人脸特征准确度，关闭可以提升性能

examples/face-example/src/test/python/model.py

@@ -0,0 +1,126 @@
+#!/usr/bin/env python
+#
+# Copyright 2023 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You may not use this file
+# except in compliance with the License. A copy of the License is located at
+#
+# http://aws.amazon.com/apache2.0/
+#
+# or in the "LICENSE.txt" file accompanying this file. This file is distributed on an "AS IS"
+# BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, express or implied. See the License for
+# the specific language governing permissions and limitations under the License.
+"""
+PyTorch resnet18 pre/post processing example.
+"""
+
+import json
+import logging
+import os
+from typing import Optional, Any
+import sklearn
+
+import torch
+import torch.nn.functional as F
+from torchvision import transforms
+
+from djl_python import Input
+from djl_python import Output
+
+
+class Processing(object):
+
+    def __init__(self):
+        self.topK = 5
+        self.image_processing = None
+        self.mapping = None
+        self.initialized = False
+
+    def initialize(self, properties: dict):
+        """
+        Initialize model.
+        """
+        self.image_processing = transforms.Compose([
+            transforms.Resize(112),
+            transforms.CenterCrop(112),
+            transforms.ToTensor(),
+            transforms.Normalize(mean=[0.485, 0.456, 0.406],
+                                 std=[0.229, 0.224, 0.225])
+        ])
+        #self.mapping = self.load_label_mapping("index_to_name.json")
+        self.initialized = True
+
+    def preprocess(self, inputs: Input) -> Output:
+        outputs = Output()
+        try:
+            batch = inputs.get_batches()
+            images = []
+            for i, item in enumerate(batch):
+                image = self.image_processing(item.get_as_image())
+                images.append(image)
+            images = torch.stack(images)
+            outputs.add_as_numpy(images.detach().numpy())
+            outputs.add_property("content-type", "tensor/ndlist")
+        except Exception as e:
+            logging.exception("pre-process failed")
+            # error handling
+            outputs = Output().error(str(e))
+
+        return outputs
+
+    def postprocess(self, inputs: Input) -> Output:
+        outputs = Output()
+        try:
+            data = inputs.get_as_numpy(0)[0]
+            item = torch.from_numpy(data)
+            print("data shape:", item.shape)
+            embedding = sklearn.preprocessing.normalize(item).flatten()
+            outputs.add(embedding)
+        except Exception as e:
+            logging.exception("post-process failed")
+            # error handling
+            outputs = Output().error(str(e))
+
+        return outputs
+
+    @staticmethod
+    def load_label_mapping(mapping_file_path: Any) -> dict:
+        if not os.path.isfile(mapping_file_path):
+            raise Exception('mapping file not found: ' + mapping_file_path)
+
+        with open(mapping_file_path) as f:
+            mapping = json.load(f)
+        if not isinstance(mapping, dict):
+            raise Exception('mapping file should be in "class":"label" format')
+
+        for key, value in mapping.items():
+            new_value = value
+            if isinstance(new_value, list):
+                new_value = value[-1]
+            if not isinstance(new_value, str):
+                raise Exception(
+                    'labels in mapping must be either str or [str]')
+            mapping[key] = new_value
+        return mapping
+
+
+_service = Processing()
+
+
+def preprocess(inputs: Input) -> Output:
+    return _service.preprocess(inputs)
+
+
+def postprocess(inputs: Input) -> Output:
+    return _service.postprocess(inputs)
+
+
+def handle(inputs: Input) -> Optional[Output]:
+    """
+    Default handler function
+    """
+    if not _service.initialized:
+        # stateful model
+        _service.initialize(inputs.get_properties())
+
+    return None

examples/objectdetection-example/pom.xml

@@ -12,7 +12,7 @@
         <maven.compiler.source>11</maven.compiler.source>
         <maven.compiler.target>11</maven.compiler.target>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
-        <smartjavaai.version>1.0.23</smartjavaai.version>
+        <smartjavaai.version>1.0.24</smartjavaai.version>
         <!--如果打包运行，需要替换成你的main-->
         <exec.mainClass>smartai.examples.objectdetection.ObjectDetection</exec.mainClass>
 
@@ -34,7 +34,7 @@
         <dependencies>
             <dependency>
                 <groupId>cn.smartjavaai</groupId>
-                <artifactId>smartjavaai-bom</artifactId>
+                <artifactId>bom</artifactId>
                 <version>${smartjavaai.version}</version>
                 <type>pom</type>
                 <!-- 注意这里是import -->
@@ -94,7 +94,7 @@
         <!--目标检测模块-->
         <dependency>
             <groupId>cn.smartjavaai</groupId>
-            <artifactId>smartjavaai-objectdetection</artifactId>
+            <artifactId>vision</artifactId>
         </dependency>
 
 

examples/objectdetection-example/src/main/java/smartai/examples/objectdetection/ObjectDetection.java

@@ -2,6 +2,9 @@ package smartai.examples.objectdetection;
 
 import ai.djl.Application;
 import ai.djl.MalformedModelException;
+import ai.djl.ModelException;
+import ai.djl.inference.Predictor;
+import ai.djl.modality.Classifications;
 import ai.djl.modality.cv.Image;
 import ai.djl.modality.cv.ImageFactory;
 import ai.djl.modality.cv.output.*;
@@ -11,6 +14,7 @@ import ai.djl.repository.zoo.ModelNotFoundException;
 import ai.djl.repository.zoo.ModelZoo;
 import ai.djl.repository.zoo.ZooModel;
 import ai.djl.training.util.ProgressBar;
+import ai.djl.translate.TranslateException;
 import cn.smartjavaai.common.config.Config;
 import cn.smartjavaai.common.entity.DetectionInfo;
 import cn.smartjavaai.common.entity.DetectionRectangle;
@@ -42,6 +46,7 @@ import java.awt.*;
 import java.awt.image.BufferedImage;
 import java.io.File;
 import java.io.IOException;
+import java.net.URL;
 import java.nio.file.Paths;
 import java.util.*;
 import java.util.List;
@@ -64,6 +69,32 @@ public class ObjectDetection {
     //设备类型
     public static DeviceEnum device = DeviceEnum.CPU;
 
+    public static void main(String[] args) throws ModelException, TranslateException, IOException {
+        Classifications classification = predict();
+        log.info("{}", classification);
+    }
+
+
+    public static Classifications predict() throws IOException, ModelException, TranslateException {
+
+        Config.setCachePath("/Users/wenjie/smartjavaai_cache");
+        URL url = new URL("https://resources.djl.ai/images/action_dance.jpg");
+        // Use DJL PyTorch model zoo model
+        Criteria<URL, Classifications> criteria =
+                Criteria.builder()
+                        .setTypes(URL.class, Classifications.class)
+                        .optModelUrls(
+                                "djl://ai.djl.mxnet/action_recognition")
+                        .optEngine("MXNet")
+                        .optProgress(new ProgressBar())
+                        .build();
+
+        try (ZooModel<URL, Classifications> inception = criteria.loadModel();
+             Predictor<URL, Classifications> action = inception.newPredictor()) {
+            return action.predict(url);
+        }
+    }
+
     @BeforeClass
     public static void beforeAll() throws IOException {
         //修改缓存路径
@@ -95,6 +126,8 @@ public class ObjectDetection {
         try {
             DetectorModelConfig config = new DetectorModelConfig();
             config.setModelEnum(DetectorModelEnum.SSD_300_RESNET50);//检测模型，目前支持19种预置模型
+            config.setModelEnum(DetectorModelEnum.YOLOV12_OFFICIAL);
+            config.setModelPath("yolov11s");
             // 指定允许的类别
 //            config.setAllowedClasses(Arrays.asList("person"));
             //指定返回检测数量
@@ -205,6 +238,32 @@ public class ObjectDetection {
         }
     }
 
+    /**
+     * tensorflow目标检测
+     */
+    @Test
+    public void objectDetection3(){
+        try {
+            DetectorModelConfig config = new DetectorModelConfig();
+            config.setModelEnum(DetectorModelEnum.TENSORFLOW2_OFFICIAL);
+            config.setModelPath("/Users/wenjie/Documents/develop/model/tensorflow/ssd_mobilenet_v2_320x320_coco17_tpu-8");
+//            config.putCustomParam("synsetUrl", "https://raw.githubusercontent.com/tensorflow/models/master/research/object_detection/data/mscoco_label_map.pbtxt");
+//            config.putCustomParam("synsetPath", "/Users/wenjie/Downloads/mscoco_label_map.pbtxt.txt");
+            config.putCustomParam("synsetFileName", "mscoco.pbtxt");
+            // 指定允许的类别
+//            config.setAllowedClasses(Arrays.asList("person"));
+            //指定返回检测数量
+            config.setTopK(100);
+            config.setDevice(device);
+            DetectorModel detectorModel = ObjectDetectionModelFactory.getInstance().getModel(config);
+            DetectionResponse detectionResponse = detectorModel.detect("src/main/resources/dog_bike_car.jpg");
+            detectorModel.detectAndDraw("src/main/resources/dog_bike_car.jpg", "output/dog_bike_car_detect.jpg");
+            log.info("目标检测结果：{}", JSONObject.toJSONString(detectionResponse));
+        } catch (Exception e) {
+            e.printStackTrace();
+        }
+    }
+
 
     /**
      * 摄像头目标检测

examples/ocr-examples/output/table_ch2_result.html

@@ -0,0 +1,5 @@
+<style>
+table { border-collapse: collapse; }
+td, th, table { border: 1px solid black; padding: 5px; }
+</style>
+<html><body><table><thead><tr><td>主要财务比率</td><td>2020</td><td>2021</td><td>2022E</td><td>2023E</td><td>2024E</td></tr></thead><tbody><tr><td>成长能力</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>营业收入</td><td>97.08%</td><td>33.28%</td><td>65.00%</td><td>42.10%</td><td>21.00%</td></tr><tr><td>营业利润</td><td>165.21%</td><td>22.38%</td><td>31.65%</td><td>64.55%</td><td>36.68%</td></tr><tr><td>归属於母公司净利润</td><td>164.75%</td><td>24.17%</td><td>39.44%</td><td>64.13%</td><td>38.63%</td></tr><tr><td>获利能力</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>毛利率</td><td>25.45%</td><td>23.01%</td><td>16.80%</td><td>17.00%</td><td>18.00%</td></tr><tr><td>净利率</td><td>13.98%</td><td>13.03%</td><td>11.01%</td><td>12.72%</td><td>14.57%</td></tr><tr><td>ROE</td><td>19.29%</td><td>19.25%</td><td>20.77%</td><td>47.11%</td><td>35.24%</td></tr><tr><td>ROIC</td><td>44.53%</td><td>41.55%</td><td>44.21%</td><td>32.59%</td><td>62.14%</td></tr><tr><td>偿债能力</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>资产负债率</td><td>48.28%</td><td>54.90%</td><td>57.79%</td><td>65.62%</td><td>58.84%</td></tr><tr><td>净负债率</td><td>-39.12%</td><td>-36.03%</td><td>6.62%</td><td>8.70%</td><td>5.28%</td></tr><tr><td>流动比率</td><td>1.77</td><td>1.74</td><td>1.60</td><td>1.41</td><td>1.65</td></tr><tr><td>速动比率</td><td>1.26</td><td>1.07</td><td>0.85</td><td>0.62</td><td>0.81</td></tr><tr><td>营运能力</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>应收账款周转率</td><td>5.16</td><td>4.59</td><td>4.11</td><td>5.24</td><td>5.24</td></tr><tr><td>存货周转率</td><td>3.48</td><td>2.89</td><td>2.55</td><td>2.77</td><td>2.63</td></tr><tr><td>总资产周转率</td><td>0.80</td><td>0.78</td><td>0.93</td><td>1.21</td><td>1.22</td></tr><tr><td>每股指标（元）</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>每股收益</td><td>0.84</td><td>1.04</td><td>1.45</td><td>2.38</td><td>3.30</td></tr><tr><td>每股经营现金流</td><td>0.03</td><td>0.04</td><td>-2.54</td><td>4.28</td><td>-1.13</td></tr><tr><td>每股净资产</td><td>4.34</td><td>5.40</td><td>6.97</td><td>5.05</td><td>9.35</td></tr><tr><td>估值比率</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>市盈率</td><td>41.30</td><td>33.26</td><td>23.85</td><td>14.53</td><td>10.48</td></tr><tr><td>市净率</td><td>7.97</td><td>6.40</td><td>4.95</td><td>6.85</td><td>3.69</td></tr><tr><td>EV/EBITDA</td><td>5.08</td><td>22.72</td><td>23.65</td><td>14.40</td><td>10.60</td></tr><tr><td>EV/EBIT</td><td>5.33</td><td>24.19</td><td>25.45</td><td>15.05</td><td>10.95</td></tr></tbody></table></body></html>