Merge pull request #69 from litongjava/master

release to v2.2.9
AnyListen · Dec 10, 2023 · 235a6cd · 235a6cd
2 parents 2e3820d + 6d57eba
commit 235a6cd
Show file tree

Hide file tree

Showing 17 changed files with 379 additions and 222 deletions.
diff --git a/docs/cn/1 快速上手/1.0 快速上手.md b/docs/cn/1 快速上手/1.0 快速上手.md
@@ -92,7 +92,7 @@ mvn clean package -DskipTests
 以下是一个基本的 `jpackage` 命令示例，用于将 JavaFX JAR 打包为 `.exe` 文件：
 
 ```bash
-jpackage --type exe --input target/ --main-jar tool-ocr-1.2.6.jar --name tree-hole-ocr --main-class com.luooqi.ocr.MainFm
+jpackage --type exe --input target/ --main-jar tool-ocr-1.2.6.jar --name tree-hole-ocr --main-class com.luooqi.ocr.OcrApp
 ```
 
 其中：

diff --git a/pom.xml b/pom.xml
@@ -6,14 +6,14 @@
 
   <groupId>com.luooqi</groupId>
   <artifactId>tools-ocr</artifactId>
-  <version>2.2.8</version>
+  <version>2.2.9</version>
 
   <properties>
     <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
     <java.version>1.8</java.version>
     <maven.compiler.source>${java.version}</maven.compiler.source>
     <maven.compiler.target>${java.version}</maven.compiler.target>
-    <main.class>com.luooqi.ocr.MainFm</main.class>
+    <main.class>com.luooqi.ocr.OcrApp</main.class>
     <lombok.version>1.18.20</lombok.version>
     <logback.version>1.2.3</logback.version>
     <djl.version>0.25.0</djl.version>
@@ -39,6 +39,12 @@
       <version>4.2</version>
     </dependency>
 
+    <dependency>
+      <groupId>org.apache.pdfbox</groupId>
+      <artifactId>pdfbox</artifactId>
+      <version>2.0.24</version>
+    </dependency>
+
     <dependency>
       <groupId>org.projectlombok</groupId>
       <artifactId>lombok</artifactId>
@@ -96,21 +102,21 @@
     </dependency>
     <!-- GraalVM - Java Fx -->
     <!-- https://mvnrepository.com/artifact/org.openjfx/javafx-controls -->
-<!--    <dependency>-->
-<!--      <groupId>org.openjfx</groupId>-->
-<!--      <artifactId>javafx-controls</artifactId>-->
-<!--      <version>17</version>-->
-<!--    </dependency>-->
-<!--    <dependency>-->
-<!--      <groupId>org.openjfx</groupId>-->
-<!--      <artifactId>javafx-swing</artifactId>-->
-<!--      <version>17</version>-->
-<!--    </dependency>-->
-<!--    <dependency>-->
-<!--      <groupId>org.openjfx</groupId>-->
-<!--      <artifactId>javafx-fxml</artifactId>-->
-<!--      <version>17</version>-->
-<!--    </dependency>-->
+    <!--    <dependency>-->
+    <!--      <groupId>org.openjfx</groupId>-->
+    <!--      <artifactId>javafx-controls</artifactId>-->
+    <!--      <version>17</version>-->
+    <!--    </dependency>-->
+    <!--    <dependency>-->
+    <!--      <groupId>org.openjfx</groupId>-->
+    <!--      <artifactId>javafx-swing</artifactId>-->
+    <!--      <version>17</version>-->
+    <!--    </dependency>-->
+    <!--    <dependency>-->
+    <!--      <groupId>org.openjfx</groupId>-->
+    <!--      <artifactId>javafx-fxml</artifactId>-->
+    <!--      <version>17</version>-->
+    <!--    </dependency>-->
 
 
     <!--windows gpu-->

diff --git a/readme.md b/readme.md
@@ -8,6 +8,9 @@
 - onnx
 - paddle ocr
 - opencv
+
+## 开源地址
+[gitee](https://gitee.com/ppnt/tools-ocr) | [github](https://github.com/litongjava/tools-ocr)
 ## 安装
 > - **安装路径请勿包含中文字符**；
 > - 本程序使用 JavaFX 开发，提供的安装包中已经包含了Java
@@ -49,6 +52,7 @@ mvn jfx:native -DskipTests -f pom.xml
 ![2](readme_files/2.jpg)
 
 ## TODO
+- [x] PDF识别
 - [x] 图片文字识别
   - [x] 识别结果文本对齐（暂未实现多分栏）
   - [x] 全屏模式下截图

diff --git a/src/main/java/com/litongjava/djl/paddle/ocr/v4/detection/OCRDetectionTranslator.java b/src/main/java/com/litongjava/djl/paddle/ocr/v4/detection/OCRDetectionTranslator.java
@@ -104,18 +104,20 @@ public NDList processOutput(TranslatorContext ctx, NDList list) {
       srcMat.release();
     }
 
+    NDList dt_boxes = null;
     NDArray boxes = boxes_from_bitmap(manager, pred, newMask);
+    if (boxes != null) {
+      //boxes[:, :, 0] = boxes[:, :, 0] / ratio_w
+      NDArray boxes1 = boxes.get(":, :, 0").div(ratio_w);
+      boxes.set(new NDIndex(":, :, 0"), boxes1);
+      //boxes[:, :, 1] = boxes[:, :, 1] / ratio_h
+      NDArray boxes2 = boxes.get(":, :, 1").div(ratio_h);
+      boxes.set(new NDIndex(":, :, 1"), boxes2);
 
-    //boxes[:, :, 0] = boxes[:, :, 0] / ratio_w
-    NDArray boxes1 = boxes.get(":, :, 0").div(ratio_w);
-    boxes.set(new NDIndex(":, :, 0"), boxes1);
-    //boxes[:, :, 1] = boxes[:, :, 1] / ratio_h
-    NDArray boxes2 = boxes.get(":, :, 1").div(ratio_h);
-    boxes.set(new NDIndex(":, :, 1"), boxes2);
+      dt_boxes = this.filter_tag_det_res(boxes);
 
-    NDList dt_boxes = this.filter_tag_det_res(boxes);
-
-    dt_boxes.detach();
+      dt_boxes.detach();
+    }
 
     // release Mat
     newMask.release();
@@ -255,12 +257,18 @@ private NDArray boxes_from_bitmap(NDManager manager, NDArray pred, Mat bitmap) {
       newContour.release();
     }
 
-    NDArray boxes = NDArrays.stack(boxList);
-
     // release
     hierarchy.release();
 
+    NDArray boxes = null;
+    if (boxList.size() > 0) {
+      boxes = NDArrays.stack(boxList);
+      return boxes;
+    }
+
     return boxes;
+
+
   }
 
   /**

diff --git a/src/main/java/com/litongjava/djl/paddle/ocr/v4/recognition/OcrV4Recognition.java b/src/main/java/com/litongjava/djl/paddle/ocr/v4/recognition/OcrV4Recognition.java
@@ -64,6 +64,9 @@ public List<RotatedBox> predict(NDManager manager,
                                   Image image, Predictor<Image, NDList> detector, Predictor<Image, String> recognizer)
     throws TranslateException {
     NDList boxes = detector.predict(image);
+    if (boxes == null) {
+      return null;
+    }
     // 交给 NDManager自动管理内存
     // attach to manager for automatic memory management
     boxes.attach(manager);

diff --git a/src/main/java/com/luooqi/ocr/OcrApp.java b/src/main/java/com/luooqi/ocr/OcrApp.java
@@ -0,0 +1,40 @@
+package com.luooqi.ocr;
+
+import cn.hutool.core.thread.GlobalThreadPool;
+import com.luooqi.ocr.config.InitConfig;
+import com.luooqi.ocr.local.PaddlePaddleOCRV4;
+import com.luooqi.ocr.windows.MainForm;
+import javafx.application.Application;
+import javafx.stage.Stage;
+import lombok.extern.slf4j.Slf4j;
+import org.jnativehook.GlobalScreen;
+
+@Slf4j
+public class OcrApp extends Application {
+
+  public static void main(String[] args) {
+    launch(args);
+  }
+
+  @Override
+  public void init() throws Exception {
+    super.init();
+    InitConfig.init();
+  }
+
+
+  @Override
+  public void start(Stage primaryStage) {
+    MainForm mainForm = new MainForm();
+    mainForm.init(primaryStage);
+    primaryStage.show();
+  }
+
+  @Override
+  public void stop() throws Exception {
+    log.info("close");
+    GlobalScreen.unregisterNativeHook();
+    PaddlePaddleOCRV4.INSTANCE.close();
+    GlobalThreadPool.shutdown(true);
+  }
+}
diff --git a/src/main/java/com/luooqi/ocr/constants/ImagesConstants.java b/src/main/java/com/luooqi/ocr/constants/ImagesConstants.java
@@ -0,0 +1,8 @@
+package com.luooqi.ocr.constants;
+
+/**
+ * Created by [email protected] on 12/9/2023_7:14 PM
+ */
+public class ImagesConstants {
+  public static final String LOGO = "img/logo.png";
+}
diff --git a/src/main/java/com/luooqi/ocr/local/LocalOCR.java b/src/main/java/com/luooqi/ocr/local/LocalOCR.java
diff --git a/src/main/java/com/luooqi/ocr/local/PaddlePaddleOCRV4.java b/src/main/java/com/luooqi/ocr/local/PaddlePaddleOCRV4.java
@@ -9,6 +9,7 @@
 import ai.djl.repository.zoo.ModelNotFoundException;
 import ai.djl.repository.zoo.ModelZoo;
 import ai.djl.repository.zoo.ZooModel;
+import ai.djl.translate.TranslateException;
 import com.litongjava.djl.paddle.ocr.v4.common.RotatedBox;
 import com.litongjava.djl.paddle.ocr.v4.common.RotatedBoxCompX;
 import com.litongjava.djl.paddle.ocr.v4.detection.OcrV4Detection;
@@ -62,7 +63,14 @@ public void init() {
   public String ocr(File imageFile) throws Exception {
     Path path = imageFile.toPath();
     Image image = OpenCVImageFactory.getInstance().fromFile(path);
+    return ocr(image);
+  }
+
+  public String ocr(Image image) throws Exception {
     List<RotatedBox> detections = recognition.predict(manager, image, detector, recognizer);
+    if (detections == null) {
+      return null;
+    }
 
     List<RotatedBox> initList = new ArrayList<>();
     for (RotatedBox result : detections) {
@@ -105,4 +113,9 @@ public String ocr(File imageFile) throws Exception {
     }
     return fullText.toString();
   }
+
+  public void close() {
+    detector.close();
+    recognizer.close();
+  }
 }
diff --git a/src/main/java/com/luooqi/ocr/snap/ScreenCapture.java b/src/main/java/com/luooqi/ocr/snap/ScreenCapture.java
@@ -3,9 +3,9 @@
 
 import cn.hutool.core.swing.ScreenUtil;
 import cn.hutool.log.StaticLog;
-import com.luooqi.ocr.MainFm;
 import com.luooqi.ocr.model.CaptureInfo;
 import com.luooqi.ocr.utils.CommUtils;
+import com.luooqi.ocr.windows.MainForm;
 import javafx.animation.AnimationTimer;
 import javafx.application.Platform;
 import javafx.embed.swing.SwingFXUtils;
@@ -180,8 +180,7 @@ public ScreenCapture(Stage mainStage) {
 
     mainCanvas.setOnMouseDragged(m -> {
       if (m.getButton() == MouseButton.PRIMARY) {
-        if (m.getScreenX() >= CaptureInfo.ScreenMinX &&
-          m.getScreenX() <= CaptureInfo.ScreenMaxX) {
+        if (m.getScreenX() >= CaptureInfo.ScreenMinX && m.getScreenX() <= CaptureInfo.ScreenMaxX) {
           data.mouseXNow = (int) m.getX();
         } else if (m.getScreenX() > CaptureInfo.ScreenMaxX) {
           data.mouseXNow = CaptureInfo.ScreenWidth;
@@ -311,8 +310,7 @@ private void addKeyHandlers() {
       }
     });
 
-    data.anyPressed.addListener((obs, wasPressed, isNowPressed) ->
-    {
+    data.anyPressed.addListener((obs, wasPressed, isNowPressed) -> {
       if (isNowPressed) {
         yPressedAnimation.start();
       } else {
@@ -370,15 +368,17 @@ private void repaintCanvas() {
         : data.mouseYNow // UP
     ;
 
-    gc.strokeRect(data.rectUpperLeftX - 1.00, data.rectUpperLeftY - 1.00, data.rectWidth + 2.00, data.rectHeight + 2.00);
+    gc.strokeRect(data.rectUpperLeftX - 1.00, data.rectUpperLeftY - 1.00, data.rectWidth + 2.00,
+      data.rectHeight + 2.00);
     gc.clearRect(data.rectUpperLeftX, data.rectUpperLeftY, data.rectWidth, data.rectHeight);
 
     // draw the text
     if (!data.hideExtraFeatures.getValue() && (data.rectWidth > 0 || data.rectHeight > 0)) {
       double middle = data.rectUpperLeftX + data.rectWidth / 2.00;
       gc.setLineWidth(1);
       gc.setFill(Color.FIREBRICK);
-      gc.fillRect(middle - 77, data.rectUpperLeftY < 50 ? data.rectUpperLeftY + 2 : data.rectUpperLeftY - 18.00, 100, 18);
+      gc.fillRect(middle - 77, data.rectUpperLeftY < 50 ? data.rectUpperLeftY + 2 : data.rectUpperLeftY - 18.00, 100,
+        18);
       gc.setFill(Color.WHITE);
       gc.fillText(data.rectWidth + " * " + data.rectHeight, middle - 77 + 9,
         data.rectUpperLeftY < 50 ? data.rectUpperLeftY + 17.00 : data.rectUpperLeftY - 4.00);
@@ -398,16 +398,16 @@ private void selectWholeScreen() {
 
   public void prepareForCapture() {
     isSnapping = true;
-    MainFm.stage.setOpacity(0.0f);
+    MainForm.stage.setOpacity(0.0f);
     Platform.runLater(() -> {
-      Rectangle rectangle = CommUtils.getDisplayScreen(MainFm.stage);
+      Rectangle rectangle = CommUtils.getDisplayScreen(MainForm.stage);
       data.reset();
       CaptureInfo.ScreenMinX = rectangle.x;
       CaptureInfo.ScreenMaxX = rectangle.x + rectangle.width;
       CaptureInfo.ScreenWidth = rectangle.width;
       CaptureInfo.ScreenHeight = rectangle.height;
       BufferedImage bufferedImage = ScreenUtil.captureScreen(rectangle);
-      //bufferedImage = Scalr.resize(bufferedImage, Scalr.Method.QUALITY, Scalr.Mode.AUTOMATIC, CaptureInfo.ScreenWidth * 2, CaptureInfo.ScreenHeight * 2);
+      // bufferedImage = Scalr.resize(bufferedImage, Scalr.Method.QUALITY, Scalr.Mode.AUTOMATIC, CaptureInfo.ScreenWidth * 2, CaptureInfo.ScreenHeight * 2);
       WritableImage fxImage = SwingFXUtils.toFXImage(bufferedImage, null);
       deActivateAllKeys();
       scene.setRoot(new Pane());
@@ -417,9 +417,9 @@ public void prepareForCapture() {
       mainCanvas.setHeight(CaptureInfo.ScreenHeight);
       mainCanvas.setCursor(Cursor.CROSSHAIR);
       initGraphContent();
-      rootPane.setBackground(new Background(new BackgroundImage(fxImage,
-        BackgroundRepeat.NO_REPEAT, BackgroundRepeat.NO_REPEAT,
-        BackgroundPosition.CENTER, new BackgroundSize(CaptureInfo.ScreenWidth, CaptureInfo.ScreenHeight, false, false, true, true))));
+      rootPane.setBackground(new Background(new BackgroundImage(fxImage, BackgroundRepeat.NO_REPEAT,
+        BackgroundRepeat.NO_REPEAT, BackgroundPosition.CENTER,
+        new BackgroundSize(CaptureInfo.ScreenWidth, CaptureInfo.ScreenHeight, false, false, true, true))));
       repaintCanvas();
       stage.setScene(scene);
       stage.setFullScreenExitHint("");
@@ -438,19 +438,21 @@ private void prepareImage() {
     BufferedImage image;
     try {
       mainCanvas.setDisable(true);
-      image = new Robot().createScreenCapture(new Rectangle(data.rectUpperLeftX + CaptureInfo.ScreenMinX, data.rectUpperLeftY + (int) CommUtils.getCrtScreen(stage).getVisualBounds().getMinY(), data.rectWidth, data.rectHeight));
+      image = new Robot().createScreenCapture(new Rectangle(data.rectUpperLeftX + CaptureInfo.ScreenMinX,
+        data.rectUpperLeftY + (int) CommUtils.getCrtScreen(stage).getVisualBounds().getMinY(), data.rectWidth,
+        data.rectHeight));
     } catch (AWTException ex) {
       StaticLog.error(ex);
       return;
     } finally {
       mainCanvas.setDisable(false);
-      MainFm.restore(false);
+      MainForm.restore(false);
     }
-    MainFm.doOcr(image);
+    MainForm.doOcr(image);
   }
 
   public void cancelSnap() {
     deActivateAllKeys();
-    MainFm.restore(true);
+    MainForm.restore(true);
   }
 }