diff --git a/model_zoo/official/lite/posenet/images/posenet_detection.png b/model_zoo/official/lite/posenet/images/posenet_detection.png index 4a3f7437fc..db253e597c 100644 Binary files a/model_zoo/official/lite/posenet/images/posenet_detection.png and b/model_zoo/official/lite/posenet/images/posenet_detection.png differ diff --git a/model_zoo/official/lite/style_transfer/README.md b/model_zoo/official/lite/style_transfer/README.md new file mode 100644 index 0000000000..4825e21436 --- /dev/null +++ b/model_zoo/official/lite/style_transfer/README.md @@ -0,0 +1,305 @@ +# MindSpore Lite 端侧骨骼检测demo（Android） + +本示例程序演示了如何在端侧利用MindSpore Lite API以及MindSpore Lite风格迁移模型完成端侧推理，根据demo内置的标准图片更换目标图片的艺术风格，并在App图像预览界面中显示出来。 + +## 运行依赖 + +- Android Studio >= 3.2 (推荐4.0以上版本) +- NDK 21.3 +- CMake 3.10 +- Android SDK >= 26 + +## 构建与运行 + +1. 在Android Studio中加载本示例源码，并安装相应的SDK（指定SDK版本后，由Android Studio自动安装）。 + + ![start_home](images/home.png) + + 启动Android Studio后，点击`File->Settings->System Settings->Android SDK`，勾选相应的SDK。如下图所示，勾选后，点击`OK`，Android Studio即可自动安装SDK。 + + ![start_sdk](images/sdk_management.png) + + 使用过程中若出现Android Studio配置问题，可参考第5项解决。 + +2. 连接Android设备，运行骨应用程序。 + + 通过USB连接Android设备调试，点击`Run 'app'`即可在你的设备上运行本示例项目。 + > 编译过程中Android Studio会自动下载MindSpore Lite、模型文件等相关依赖项，编译过程需做耐心等待。 + + ![run_app](images/run_app.PNG) + + Android Studio连接设备调试操作，可参考。 + +3. 在Android设备上，点击“继续安装”，安装完即可查看到推理结果。 + + ![install](images/install.jpg) + + 使用风格迁移demo时，用户可先导入或拍摄自己的图片，然后选择一种预置风格，得到推理后的新图片，最后使用还原或保存新图片功能。 + + 原始图片： + + ![sult](images/style_transfer_demo.png) + + 风格迁移后的新图片： + + ![sult](images/style_transfer_result.png) + +4. Android Studio 配置问题解决方案可参考下表： + + | | 报错 | 解决方案 | + | ---- | ------------------------------------------------------------ | ------------------------------------------------------------ | + | 1 | Gradle sync failed: NDK not configured. | 在local.properties中指定安装的ndk目录：ndk.dir={ndk的安装目录} | + | 2 | Requested NDK version did not match the version requested by ndk.dir | 可手动下载相应的[NDK版本](https://developer.android.com/ndk/downloads?hl=zh-cn)，并在Project Structure - Android NDK location设置中指定SDK的位置（可参考下图完成） | + | 3 | This version of Android Studio cannot open this project, please retry with Android Studio or newer. | 在工具栏-help-Checkout for Updates中更新版本 | + | 4 | SSL peer shut down incorrectly | 重新构建 | + + ![project_structure](images/project_structure.png) + +## 示例程序详细说明 + +风格Android示例程序通过Android Camera 2 API实现摄像头获取图像帧，以及相应的图像处理等功能，在[Runtime](https://www.mindspore.cn/tutorial/lite/zh-CN/master/use/runtime.html)中完成模型推理的过程。 + +### 示例程序结构 + +```text + +├── app +│   ├── build.gradle # 其他Android配置文件 +│   ├── download.gradle # APP构建时由gradle自动从HuaWei Server下载依赖的库文件及模型文件 +│   ├── proguard-rules.pro +│   └── src +│   ├── main +│   │   ├── AndroidManifest.xml # Android配置文件 +│   │   ├── java # java层应用代码 +│   │   │   └── com +│   │   │   └── mindspore +│   │   │   └── posenetdemo # 图像处理及推理流程实现 +│   │   │   ├── CameraDataDealListener.java +│   │   │   ├── ImageUtils.java +│   │   │   ├── MainActivity.java +│   │   │   ├── PoseNetFragment.java +│   │   │   ├── Posenet.java # +│   │   │   └── TestActivity.java +│   │   └── res # 存放Android相关的资源文件 +│   └── test +└── ... + +``` + +### 下载及部署模型文件 + +从MindSpore Model Hub中下载模型文件，本示例程序中使用的目标检测模型文件为`posenet_model.ms`，同样通过`download.gradle`脚本在APP构建时自动下载，并放置在`app/src/main/assets`工程目录下。 + +> 若下载失败请手动下载模型文件，style_predict_quant.ms [下载链接](https://download.mindspore.cn/model_zoo/official/lite/style_lite/style_predict_quant.ms)，以及style_transfer_quant.ms [下载链接](https://download.mindspore.cn/model_zoo/official/lite/style_lite/style_transfer_quant.ms)。 + +### 编写端侧推理代码 + +在风格迁移demo中，使用Java API实现端测推理。相比于C++ API，Java API可以直接在Java Class中调用，无需实现JNI层的相关代码，具有更好的便捷性。 + +风格迁移demo推理代码流程如下，完整代码请参见：`src/main/java/com/mindspore/styletransferdemo/StyleTransferModelExecutor.java`。 + +1. 加载MindSpore Lite模型文件，构建上下文、会话以及用于推理的计算图。 + + - 加载模型：从文件系统中读取MindSpore Lite模型，并进行模型解析。 + + ```java + // Load the .ms model. + style_predict_model = new Model(); + if (!style_predict_model.loadModel(mContext, "style_predict_quant.ms")) { + Log.e("MS_LITE", "Load style_predict_model failed"); + } + + style_transform_model = new Model(); + if (!style_transform_model.loadModel(mContext, "style_transfer_quant.ms")) { + Log.e("MS_LITE", "Load style_transform_model failed"); + } + ``` + + - 创建配置上下文：创建配置上下文`MSConfig`，保存会话所需的一些基本配置参数，用于指导图编译和图执行。 + + ```java + msConfig = new MSConfig(); + if (!msConfig.init(DeviceType.DT_CPU, NUM_THREADS, CpuBindMode.MID_CPU)) { + Log.e("MS_LITE", "Init context failed"); + } + ``` + + - 创建会话：创建`LiteSession`，并调用`init`方法将上一步得到`MSConfig`配置到会话中。 + + ```java + // Create the MindSpore lite session. + Predict_session = new LiteSession(); + if (!Predict_session.init(msConfig)) { + Log.e("MS_LITE", "Create Predict_session failed"); + msConfig.free(); + } + + Transform_session = new LiteSession(); + if (!Transform_session.init(msConfig)) { + Log.e("MS_LITE", "Create Predict_session failed"); + msConfig.free(); + } + msConfig.free(); + ``` + + - 加载模型文件并构建用于推理的计算图 + + ```java + // Complile graph. + if (!Predict_session.compileGraph(style_predict_model)) { + Log.e("MS_LITE", "Compile style_predict graph failed"); + style_predict_model.freeBuffer(); + } + if (!Transform_session.compileGraph(style_transform_model)) { + Log.e("MS_LITE", "Compile style_transform graph failed"); + style_transform_model.freeBuffer(); + } + + // Note: when use model.freeBuffer(), the model can not be complile graph again. + style_predict_model.freeBuffer(); + style_transform_model.freeBuffer(); + ``` + +2. 输入数据： Java目前支持`byte[]`或者`ByteBuffer`两种类型的数据，设置输入Tensor的数据。 + + - 在输入数据之前，将float数组转换为byte数组。 + + ```java + + public static byte[] floatArrayToByteArray(float[] floats) { + ByteBuffer buffer = ByteBuffer.allocate(4 * floats.length); + buffer.order(ByteOrder.nativeOrder()); + FloatBuffer floatBuffer = buffer.asFloatBuffer(); + floatBuffer.put(floats); + return buffer.array(); + } + ``` + + - 通过`ByteBuffer`输入数据。`contentImage`为用户提供的图片，`styleBitmap`为预置风格图片。 + + ```java + public ModelExecutionResult execute(Bitmap contentImage, Bitmap styleBitmap) { + Log.i(TAG, "running models"); + fullExecutionTime = SystemClock.uptimeMillis(); + preProcessTime = SystemClock.uptimeMillis(); + ByteBuffer contentArray = + ImageUtils.bitmapToByteBuffer(contentImage, CONTENT_IMAGE_SIZE, CONTENT_IMAGE_SIZE, 0, 255); + ByteBuffer input = ImageUtils.bitmapToByteBuffer(styleBitmap, STYLE_IMAGE_SIZE, STYLE_IMAGE_SIZE, 0, 255); + ``` + +3. 对输入Tensor按照模型进行推理，获取输出Tensor，并进行后处理。 + + - 使用`runGraph`对预置图片进行模型推理，并获取结果`Predict_results`。 + + ```java + List Predict_inputs = Predict_session.getInputs(); + if (Predict_inputs.size() != 1) { + return null; + } + MSTensor Predict_inTensor = Predict_inputs.get(0); + Predict_inTensor.setData(input); + + preProcessTime = SystemClock.uptimeMillis() - preProcessTime; + stylePredictTime = SystemClock.uptimeMillis(); + + + if (!Predict_session.runGraph()) { + Log.e("MS_LITE", "Run Predict_graph failed"); + return null; + } + stylePredictTime = SystemClock.uptimeMillis() - stylePredictTime; + Log.d(TAG, "Style Predict Time to run: " + stylePredictTime); + + // Get output tensor values. + List tensorNames = Predict_session.getOutputTensorNames(); + Map outputs = Predict_session.getOutputMapByTensor(); + Set> entrys = outputs.entrySet(); + + float[] Predict_results = null; + for (String tensorName : tensorNames) { + MSTensor output = outputs.get(tensorName); + if (output == null) { + Log.e("MS_LITE", "Can not find Predict_session output " + tensorName); + return null; + } + int type = output.getDataType(); + Predict_results = output.getFloatData(); + } + ``` + + - 利用上一步获取的结果，再次对用户图片进行模型推理，得到风格转换的数据`transform_results`。 + + ```java + List Transform_inputs = Transform_session.getInputs(); + // transform model have 2 input tensor, tensor0: 1*1*1*100, tensor1；1*384*384*3 + MSTensor Transform_inputs_inTensor0 = Transform_inputs.get(0); + Transform_inputs_inTensor0.setData(floatArrayToByteArray(Predict_results)); + + MSTensor Transform_inputs_inTensor1 = Transform_inputs.get(1); + Transform_inputs_inTensor1.setData(contentArray); + + + styleTransferTime = SystemClock.uptimeMillis(); + + if (!Transform_session.runGraph()) { + Log.e("MS_LITE", "Run Transform_graph failed"); + return null; + } + + styleTransferTime = SystemClock.uptimeMillis() - styleTransferTime; + Log.d(TAG, "Style apply Time to run: " + styleTransferTime); + + postProcessTime = SystemClock.uptimeMillis(); + + // Get output tensor values. + List Transform_tensorNames = Transform_session.getOutputTensorNames(); + Map Transform_outputs = Transform_session.getOutputMapByTensor(); + + float[] transform_results = null; + for (String tensorName : Transform_tensorNames) { + MSTensor output1 = Transform_outputs.get(tensorName); + if (output1 == null) { + Log.e("MS_LITE", "Can not find Transform_session output " + tensorName); + return null; + } + transform_results = output1.getFloatData(); + } + ``` + + - 对输出节点的数据进行处理，得到推理后的最终结果。 + + ```java + float[][][][] outputImage = new float[1][][][]; // 1 384 384 3 + for (int x = 0; x < 1; x++) { + float[][][] arrayThree = new float[CONTENT_IMAGE_SIZE][][]; + for (int y = 0; y < CONTENT_IMAGE_SIZE; y++) { + float[][] arrayTwo = new float[CONTENT_IMAGE_SIZE][]; + for (int z = 0; z < CONTENT_IMAGE_SIZE; z++) { + float[] arrayOne = new float[3]; + for (int i = 0; i < 3; i++) { + int n = i + z * 3 + y * CONTENT_IMAGE_SIZE * 3 + x * CONTENT_IMAGE_SIZE * CONTENT_IMAGE_SIZE * 3; + arrayOne[i] = transform_results[n]; + } + arrayTwo[z] = arrayOne; + } + arrayThree[y] = arrayTwo; + } + outputImage[x] = arrayThree; + } + + + Bitmap styledImage = + ImageUtils.convertArrayToBitmap(outputImage, CONTENT_IMAGE_SIZE, CONTENT_IMAGE_SIZE); + postProcessTime = SystemClock.uptimeMillis() - postProcessTime; + + fullExecutionTime = SystemClock.uptimeMillis() - fullExecutionTime; + Log.d(TAG, "Time to run everything: $" + fullExecutionTime); + + return new ModelExecutionResult(styledImage, + preProcessTime, + stylePredictTime, + styleTransferTime, + postProcessTime, + fullExecutionTime, + formatExecutionLog()); + ``` diff --git a/model_zoo/official/lite/style_transfer/images/home.png b/model_zoo/official/lite/style_transfer/images/home.png new file mode 100644 index 0000000000..29e954a425 Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/home.png differ diff --git a/model_zoo/official/lite/style_transfer/images/install.jpg b/model_zoo/official/lite/style_transfer/images/install.jpg new file mode 100644 index 0000000000..c98ee71dae Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/install.jpg differ diff --git a/model_zoo/official/lite/style_transfer/images/project_structure.png b/model_zoo/official/lite/style_transfer/images/project_structure.png new file mode 100644 index 0000000000..6f71294479 Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/project_structure.png differ diff --git a/model_zoo/official/lite/style_transfer/images/run_app.PNG b/model_zoo/official/lite/style_transfer/images/run_app.PNG new file mode 100644 index 0000000000..2557b6293d Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/run_app.PNG differ diff --git a/model_zoo/official/lite/style_transfer/images/sdk_management.png b/model_zoo/official/lite/style_transfer/images/sdk_management.png new file mode 100644 index 0000000000..faf694bd2e Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/sdk_management.png differ diff --git a/model_zoo/official/lite/style_transfer/images/style_transfer_demo.png b/model_zoo/official/lite/style_transfer/images/style_transfer_demo.png new file mode 100644 index 0000000000..ba2fe024d6 Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/style_transfer_demo.png differ diff --git a/model_zoo/official/lite/style_transfer/images/style_transfer_result.png b/model_zoo/official/lite/style_transfer/images/style_transfer_result.png new file mode 100644 index 0000000000..cb066922a3 Binary files /dev/null and b/model_zoo/official/lite/style_transfer/images/style_transfer_result.png differ