机器学习套件分析器
使用集合让一切井井有条
根据您的偏好保存内容并对其进行分类。
Google 的 机器学习套件可提供设备端机器学习 Vision API,用于检测人脸、扫描条形码、为图片加标签等。借助机器学习套件分析器,您可以更轻松地将机器学习套件与 CameraX 应用集成。
机器学习套件分析器是 ImageAnalysis.Analyzer
接口的实现。该分析器会通过替换默认目标分辨率(如果需要)来针对机器学习套件的使用进行优化,处理坐标转换,并将框架传递给会返回汇总分析结果的机器学习套件。
实现机器学习套件分析器
如需实现机器学习套件分析器,建议使用 CameraController
类,该类可与 PreviewView
配合使用来显示界面元素。在使用 CameraController
进行实现时,机器学习套件分析器会为您处理原始 ImageAnalysis
流与 PreviewView
之间的坐标转换。该分析器会从 CameraX 接收目标坐标系,计算坐标转换,并将其转发给机器学习套件的 Detector
类进行分析。
若要将机器学习套件分析器与 CameraController
搭配使用,请调用 setImageAnalysisAnalyzer()
并向它传递一个新的机器学习套件分析器对象,同时在其构造函数中包含以下内容:
以下代码会使用 CameraController
来实现机器学习套件分析器,以设置用于检测二维码的 BarcodeScanner
:
Kotlin
// create BarcodeScanner object
val options = BarcodeScannerOptions.Builder()
.setBarcodeFormats(Barcode.FORMAT_QR_CODE)
.build()
val barcodeScanner = BarcodeScanning.getClient(options)
cameraController.setImageAnalysisAnalyzer(
ContextCompat.getMainExecutor(this),
MlKitAnalyzer(
listOf(barcodeScanner),
COORDINATE_SYSTEM_VIEW_REFERENCED,
ContextCompat.getMainExecutor(this)
) { result: MlKitAnalyzer.Result? ->
// The value of result.getResult(barcodeScanner) can be used directly for drawing UI overlay.
}
)
Java
// create BarcodeScanner object
BarcodeScannerOptions options = new BarcodeScannerOptions.Builder()
.setBarcodeFormats(Barcode.FORMAT_QR_CODE)
.build();
BarcodeScanner barcodeScanner = BarcodeScanning.getClient(options);
cameraController.setImageAnalysisAnalyzer(executor,
new MlKitAnalyzer(List.of(barcodeScanner), COORDINATE_SYSTEM_VIEW_REFERENCED,
executor, result -> {
// The value of result.getResult(barcodeScanner) can be used directly for drawing UI overlay.
});
在上面的代码示例中,机器学习套件分析器会将以下内容传递给 BarcodeScanner
的 Detector
类:
- 基于代表目标坐标系的
COORDINATE_SYSTEM_VIEW_REFERENCED
的转换 Matrix。
- 相机框架。
如果 BarcodeScanner
遇到任何问题,它的 Detector
会抛出错误,并且机器学习套件分析器会将该错误传播到您的应用。如果成功,机器学习套件分析器将返回 MLKitAnalyzer.Result#getValue()
,在本例中为 Barcode
对象。
您还可以使用 camera-core
中的 ImageAnalysis
类来实现机器学习套件分析器。不过,由于 ImageAnalysis
并未与 PreviewView
集成,因此您必须手动处理坐标转换。如需了解详情,请参阅机器学习套件分析器参考文档。
其他资源
如需获取具有机器学习套件分析器功能且能正常运行的相机应用,请参阅 CameraX-MLKit 示例。
本页面上的内容和代码示例受内容许可部分所述许可的限制。Java 和 OpenJDK 是 Oracle 和/或其关联公司的注册商标。
最后更新时间 (UTC):2025-07-27。
[null,null,["最后更新时间 (UTC):2025-07-27。"],[],[],null,["# ML Kit Analyzer\n\nGoogle's [ML Kit](https://developers.google.com/ml-kit/guides) provides on-device machine learning Vision APIs for detecting\nfaces, scanning barcodes, labeling images, and more. ML Kit Analyzer makes it\neasier to integrate ML Kit with your CameraX app.\n\nML Kit Analyzer is an implementation of the [`ImageAnalysis.Analyzer`](/reference/androidx/camera/core/ImageAnalysis.Analyzer) interface. It overrides the [default target resolution](/reference/androidx/camera/core/ImageAnalysis.Analyzer#getDefaultTargetResolution())\n(if needed) to optimize for ML Kit usage, handles the coordinate transformations,\nand passes the frames to ML Kit, which returns the aggregated analysis results.\n\nImplement ML Kit Analyzer\n-------------------------\n\nTo implement ML Kit Analyzer, we recommend using the [`CameraController`](/reference/androidx/camera/view/CameraController) class, which works with [`PreviewView`](/reference/androidx/camera/view/PreviewView) to display UI elements. When implemented using `CameraController`, ML Kit Analyzer\nhandles the coordinate transformations between the original `ImageAnalysis`\nstream and `PreviewView` for you. It receives the target coordinate system from\nCameraX, calculates the coordinate transformation,\nand forwards it to ML Kit's [`Detector`](https://developers.google.com/android/reference/com/google/mlkit/vision/interfaces/Detector) class for analysis.\n\nTo use ML Kit Analyzer with `CameraController`, call [`setImageAnalysisAnalyzer()`](/reference/androidx/camera/view/CameraController#setImageAnalysisAnalyzer(java.util.concurrent.Executor,androidx.camera.core.ImageAnalysis.Analyzer)) and pass it\na new ML Kit Analyzer object with the following in its constructor:\n\n- A list of ML Kit `Detector`s, which CameraX invokes sequentially in order.\n- The target coordinate system that determines the coordinates of the ML Kit output:\n\n - [`COORDINATE_SYSTEM_VIEW_REFERENCED`](/reference/androidx/camera/view/CameraController#COORDINATE_SYSTEM_VIEW_REFERENCED()): the transformed `PreviewView` coordinates.\n - [`COORDINATE_SYSTEM_ORIGINAL`](/reference/androidx/camera/core/ImageAnalysis#COORDINATE_SYSTEM_ORIGINAL()): the original `ImageAnalysis` stream coordinates.\n- An [`Executor`](/reference/java/util/concurrent/Executor) that invokes the Consumer callback and delivers\n the [`MlKitAnalyzer.Result`](/reference/androidx/camera/mlkit/vision/MlKitAnalyzer.Result), or the aggregated ML Kit result of a camera frame, to the app.\n\n- A [`Consumer`](/reference/androidx/core/util/Consumer), which CameraX invokes when there is new ML Kit output.\n\nThe following code implements ML Kit Analyzer using `CameraController` to set up\na [`BarcodeScanner`](https://developers.google.com/android/reference/com/google/mlkit/vision/barcode/BarcodeScanner) to detect QR codes: \n\n### Kotlin\n\n```kotlin\n// create BarcodeScanner object\nval options = BarcodeScannerOptions.Builder()\n .setBarcodeFormats(Barcode.FORMAT_QR_CODE)\n .build()\nval barcodeScanner = BarcodeScanning.getClient(options)\n\ncameraController.setImageAnalysisAnalyzer(\n ContextCompat.getMainExecutor(this),\n MlKitAnalyzer(\n listOf(barcodeScanner),\n COORDINATE_SYSTEM_VIEW_REFERENCED,\n ContextCompat.getMainExecutor(this)\n ) { result: MlKitAnalyzer.Result? -\u003e\n // The value of result.getResult(barcodeScanner) can be used directly for drawing UI overlay.\n }\n)\n```\n\n### Java\n\n```java\n// create BarcodeScanner object\nBarcodeScannerOptions options = new BarcodeScannerOptions.Builder()\n .setBarcodeFormats(Barcode.FORMAT_QR_CODE)\n .build();\nBarcodeScanner barcodeScanner = BarcodeScanning.getClient(options);\n\ncameraController.setImageAnalysisAnalyzer(executor,\n new MlKitAnalyzer(List.of(barcodeScanner), COORDINATE_SYSTEM_VIEW_REFERENCED,\n executor, result -\u003e {\n // The value of result.getResult(barcodeScanner) can be used directly for drawing UI overlay.\n });\n```\n\nIn the code sample above, ML Kit Analyzer passes the following to\n`BarcodeScanner`'s `Detector` class:\n\n- The transformation [Matrix](/reference/android/graphics/Matrix) based on `COORDINATE_SYSTEM_VIEW_REFERENCED` that represents the target coordinate system.\n- The camera frames.\n\nIf `BarcodeScanner` runs into any issues, then its `Detector` [throws an error](/reference/androidx/camera/mlkit/vision/MlKitAnalyzer.Result#getThrowable(com.google.mlkit.vision.interfaces.Detector%3C?%3E)),\nand ML Kit Analyzer propagates it to your app. If successful, then ML Kit Analyzer returns [`MLKitAnalyzer.Result#getValue()`](/reference/androidx/camera/mlkit/vision/MlKitAnalyzer.Result#getValue(com.google.mlkit.vision.interfaces.Detector%3CT%3E)), which\nin this case is the [`Barcode`](https://developers.google.com/android/reference/com/google/mlkit/vision/barcode/common/Barcode) object.\n\nYou can also implement ML Kit Analyzer using the [`ImageAnalysis`](/reference/androidx/camera/core/ImageAnalysis) class that is part of `camera-core`. However, because `ImageAnalysis`\nis not integrated with `PreviewView`,\nyou must manually handle the coordinate transformations. For more information,\nsee the [ML Kit Analyzer](/reference/androidx/camera/mlkit/vision/MlKitAnalyzer) reference documentation.\n\nAdditional resources\n--------------------\n\nFor a working camera app with ML Kit Analyzer functionality,\nsee the [CameraX-MLKit](https://github.com/android/camera-samples/tree/main/CameraX-MLKit) sample."]]