基于PaddleOCR文字识别Android端打包成aar包
目的
官方其实已经提供了详细的端侧部署，如果按照官方步骤无法正常编译的，可在评论区留言，我将详细编译步骤也发布出来，在这里我主要讲一下如何将官方的提供的代码转成可方便移植的module包，便于方便引入各个项目中。
官方2.5编译教程：PaddleOCR/readme.md at release/2.5 · PaddlePaddle/PaddleOCR (github.com)
开始！！！
官方包结构对比module包结构，可以发现后面结构清晰明了
1、首先将官方包下所有文件拷贝到新建的Module包下
2、删除以下文件
3、新建单例模式启动OCRPredictor类和接口OnImagePredictorListener类，代码如下：
package com.yangy.ocr;
import android.annotation.SuppressLint;
import android.content.Context;
import android.graphics.Bitmap;
import android.graphics.BitmapFactory;
import android.media.ExifInterface;
import android.os.Binder;
import android.os.Build;
import android.os.Bundle;
import android.os.Handler;
import android.os.HandlerThread;
import android.os.Message;
import android.text.TextUtils;
import android.util.Base64;
import android.util.Log;
import java.io.IOException;
public class OCRPredictor {
public final int REQUEST_LOAD_MODEL = 0;
public final int REQUEST_RUN_MODEL = 1;
public final int RESPONSE_LOAD_MODEL_SUCCESSED = 0;
public final int RESPONSE_LOAD_MODEL_FAILED = 1;
public final int RESPONSE_RUN_MODEL_SUCCESSED = 2;
public final int RESPONSE_RUN_MODEL_FAILED = 3;
public final int RESPONSE_RUN_MODEL_NO_FILE = 4;
// private final String assetModelDirPath = "models/bank";
// private final String assetlabelFilePath = "labels/ppocr_keys_bank.txt";
private final String assetModelDirPath = "models/ocr";
private final String assetlabelFilePath = "labels/ppocr_keys_v1.txt";
protected Handler sender = null; // Send command to worker thread
protected HandlerThread worker = null; // Worker thread to load&run model
protected volatile Predictor predictor = null;
private OnImagePredictorListener listener;
@SuppressLint("StaticFieldLeak")
private static OCRPredictor ocrPredictor;
private Context context;
public static OCRPredictor getInstance(Context context) {
if (ocrPredictor == null) {
ocrPredictor = new OCRPredictor(context);
}
return ocrPredictor;
}
public OCRPredictor(Context context) {
this.context = context;
init();
}
/**
*
*/
private void init() {
worker = new HandlerThread("Predictor Worker");
worker.start();
sender = new Handler(worker.getLooper()) {
public void handleMessage(Message msg) {
switch (msg.what) {
case REQUEST_LOAD_MODEL:
// Load model and reload test image
if (!onLoadModel(context)) {
predictor = null;
Log.d("OCR", "加载模型失败！");
} else {
Log.d("OCR", "加载模型成功！");
}
break;
case REQUEST_RUN_MODEL:
Bundle bundle = msg.getData();
if (predictor == null) {
receiver.sendEmptyMessage(RESPONSE_LOAD_MODEL_FAILED);
} else {
// Run model if model is loaded
if (!TextUtils.isEmpty(bundle.getString("imagePath"))) {
if (!onRunModel(bundle.getString("imagePath"))) {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_FAILED);
} else {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_SUCCESSED);
}
} else {
ImageBinder imageBinder = (ImageBinder) bundle.getBinder("bitmap");
if (imageBinder != null) {
Bitmap bitmap = imageBinder.getBitmap();
if (!onRunModel(bitmap)) {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_FAILED);
} else {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_SUCCESSED);
}
} else {
StringBinder stringBinder = (StringBinder) bundle.getBinder("base64");
if (stringBinder != null) {
String str = stringBinder.getStr();
if (!onRunModelBase64(str)) {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_FAILED);
} else {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_SUCCESSED);
}
} else {
receiver.sendEmptyMessage(RESPONSE_RUN_MODEL_NO_FILE);
}
}
}
}
break;
}
}
};
sender.sendEmptyMessage(REQUEST_LOAD_MODEL);
}
private Handler receiver = new Handler() {
@Override
public void handleMessage(Message msg) {
switch (msg.what) {
// case RESPONSE_LOAD_MODEL_SUCCESSED:
// sender.sendEmptyMessage(REQUEST_RUN_MODEL);
// break;
case RESPONSE_LOAD_MODEL_FAILED:
listener.failure(0, "加载模型失败！");
break;
case RESPONSE_RUN_MODEL_SUCCESSED:
listener.success(predictor.outputResult, predictor.outputResultList, predictor.outputImage);
break;
case RESPONSE_RUN_MODEL_FAILED:
listener.failure(1, "运行模型失败！");
break;
case RESPONSE_RUN_MODEL_NO_FILE:
listener.failure(2, "运行模型失败，请传入待识别图片！");
break;
default:
break;
}
}
};
public void predictor(final String imagePath, final OnImagePredictorListener listener) {
this.listener = listener;
Message message = new Message();
message.what = REQUEST_RUN_MODEL;
Bundle bundle = new Bundle();
bundle.putString("imagePath", imagePath);
message.setData(bundle);
sender.sendMessage(message);
}
public void predictorBitMap(final Bitmap bitmap, final OnImagePredictorListener listener) {
this.listener = listener;
Message message = new Message();
message.what = REQUEST_RUN_MODEL;
Bundle bundle = new Bundle();
bundle.putBinder("bitmap", new ImageBinder(bitmap));
message.setData(bundle);
sender.sendMessage(message);
}
public void predictorBase64(final String base64, final OnImagePredictorListener listener) {
this.listener = listener;
Message message = new Message();
message.what = REQUEST_RUN_MODEL;
Bundle bundle = new Bundle();
bundle.putBinder("base64", new StringBinder(base64));
message.setData(bundle);
sender.sendMessage(message);
}
private static class ImageBinder extends Binder {
private Bitmap bitmap;
public ImageBinder(Bitmap bitmap) {
this.bitmap = bitmap;
}
Bitmap getBitmap() {
return bitmap;
}
}
private static class StringBinder extends Binder {
private String str;
public StringBinder(String str) {
this.str = str;
}
String getStr() {
return str;
}
}
/**
* call in onCreate, model init
*
* @return
*/
private boolean onLoadModel(Context context) {
if (predictor == null) {
predictor = new Predictor();
}
return predictor.init(context, assetModelDirPath, assetlabelFilePath);
}
private boolean onRunModel(String imagePath) {
try {
ExifInterface exif = null;
exif = new ExifInterface(imagePath);
int orientation = exif.getAttributeInt(ExifInterface.TAG_ORIENTATION,
ExifInterface.ORIENTATION_UNDEFINED);
Log.i("OCR", "rotation " + orientation);
Bitmap image = BitmapFactory.decodeFile(imagePath);
image = Utils.rotateBitmap(image, orientation);
predictor.setInputImage(image);
return predictor.isLoaded() && predictor.runModel();
} catch (IOException e) {
e.printStackTrace();
return false;
}
}
private boolean onRunModel(Bitmap image) {
try {
predictor.setInputImage(image);
return predictor.isLoaded() && predictor.runModel();
} catch (Exception e) {
e.printStackTrace();
return false;
}
}
private boolean onRunModelBase64(String base64) {
try {
byte[] bitmapByte = Base64.decode(base64, Base64.DEFAULT);
Bitmap bitmap = BitmapFactory.decodeByteArray(bitmapByte, 0, bitmapByte.length);
predictor.setInputImage(bitmap);
return predictor.isLoaded() && predictor.runModel();
} catch (Exception e) {
e.printStackTrace();
return false;
}
}
/**
* 卸载模型
*/
private void onUnloadModel() {
if (predictor != null) {
predictor.releaseModel();
predictor = null;
}
}
/**
* 退出线程
*/
private void quit() {
if (worker != null) {
if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.JELLY_BEAN_MR2) {
worker.quitSafely();
} else {
worker.quit();
}
}
}
/**
* 销毁
*/
public void onDestroy() {
onUnloadModel();
quit();
ocrPredictor = null;
}
}
package com.yangy.ocr;
import android.graphics.Bitmap;
import java.util.ArrayList;
public abstract class OnImagePredictorListener {
public abstract void success(String result, ArrayList list, Bitmap bitmap);
public void failure(int code, String message) {
}
} 4、修改Predictor类，加入以下两行
5、修改native.cpp文件，将官方的_com_baidu_paddle_lite_demo_ocr_替换成我们自己的包名，如_com_yangy_ocr_，如下截图：
6、最后一步，删除Module包下build.gradle中下载PaddleLite、OpenCV包的代码，红框部分全部删除，我们不需要 ，PaddleLite和模型包需要自行编译
7、使用方式，在主App下引入该Module包，短短几行代码就可实现
完毕！！！
这里发布一个自己打的包，引入aar包后，使用如上面最后一图：
1、minSdkVersion 21
2、支持arm64-v8a和armabi-v7a
下载地址