ProcessCameraProvider

@ExperimentalCameraProviderConfiguration
fun configureInstance(cameraXConfig: CameraXConfig): Unit

Perform one-time configuration of the ProcessCameraProvider singleton with the given CameraXConfig.

This method allows configuration of the camera provider via CameraXConfig. All initialization tasks, such as communicating with the camera service, will be executed on the java.util.concurrent.Executor set by CameraXConfig.Builder.setCameraExecutor, or by an internally defined executor if none is provided.

This method is not required for every application. If the method is not called and CameraXConfig.Provider is not implemented in Application, default configuration will be used.

Once this method is called, the instance configured by the given CameraXConfig can be retrieved with getInstance. CameraXConfig.Provider implemented in Application will be ignored.

Configuration can only occur once. Once the ProcessCameraProvider has been configured with configureInstance() or getInstance, this method will throw an IllegalStateException. Because configuration can only occur once, usage of this method from library code is not recommended as the application owner should ultimately be in control of singleton configuration.

fun getInstance(context: Context): ListenableFuture<ProcessCameraProvider>

Retrieves the ProcessCameraProvider associated with the current process.

The instance returned here can be used to bind use cases to any LifecycleOwner with bindToLifecycle.

The instance's configuration may be customized by subclassing the application's Application class and implementing CameraXConfig.Provider. For example, the sample implements CameraXConfig.Provider.getCameraXConfig and initializes this process camera provider with a Camera2 implementation from androidx.camera.camera2, and with a custom executor.

import androidx.camera.camera2.Camera2Config
import androidx.camera.core.CameraXConfig

@Override
fun getCameraXConfig(): CameraXConfig {
    return CameraXConfig.Builder.fromConfig(Camera2Config.defaultConfig())
        .setCameraExecutor(executor)
        .setSchedulerHandler(handler)
        .build()
}

If it isn't possible to subclass the Application class, such as in library code, then the singleton can be configured via configureInstance before the first invocation of getInstance(context), the sample implements a customized camera provider that configures the instance before getting it.

import androidx.camera.camera2.Camera2Config
import androidx.camera.core.CameraProvider
import androidx.camera.core.CameraXConfig
import androidx.camera.lifecycle.ExperimentalCameraProviderConfiguration
import androidx.camera.lifecycle.ProcessCameraProvider
import androidx.camera.lifecycle.ProcessCameraProvider.Companion.configureInstance
import androidx.concurrent.futures.await

var configured = false // Whether the camera provider has been configured or not.

@androidx.annotation.OptIn(ExperimentalCameraProviderConfiguration::class)
suspend fun getInstance(context: Context): ProcessCameraProvider {
    synchronized(CameraProvider::class.java) {
        if (!configured) {
            configured = true
            configureInstance(
                CameraXConfig.Builder.fromConfig(Camera2Config.defaultConfig())
                    .setCameraExecutor(executor)
                    .setSchedulerHandler(scheduleHandler)
                    .build()
            )
        }
    }
    return ProcessCameraProvider.getInstance(context).await()
}

If no CameraXConfig.Provider is implemented by Application, or if the singleton has not been configured via configureInstance a default configuration will be used.

@MainThread
fun bindToLifecycle(
    singleCameraConfigs: List<ConcurrentCamera.SingleCameraConfig?>
): ConcurrentCamera

Binds list of SingleCameraConfigs to LifecycleOwner.

The concurrent camera is only supporting two cameras currently. If the input list of SingleCameraConfigs have less or more than two SingleCameraConfigs, IllegalArgumentException will be thrown. If cameras are already used by other UseCases, UnsupportedOperationException will be thrown.

A logical camera is a grouping of two or more of those physical cameras. See Multi-camera API

If we want to open concurrent logical cameras, which are one front camera and one back camera, the device needs to support PackageManager.FEATURE_CAMERA_CONCURRENT. To set up concurrent logical camera, call availableConcurrentCameraInfos to get the list of available combinations of concurrent cameras. Each sub-list contains the CameraInfos for a combination of cameras that can be operated concurrently. Each logical camera can have its own UseCases and LifecycleOwner. See {@docRoot}training/camerax/architecture#lifecycles

If the concurrent logical cameras are binding the same preview and video capture use cases, the concurrent cameras video recording will be supported. The concurrent camera preview stream will be shared with video capture and record the concurrent cameras streams as a composited stream. The CompositionSettings can be used to configure the position of each camera stream and different layouts can be built. See CompositionSettings for more details.

If we want to open concurrent physical cameras, which are two front cameras or two back cameras, the device needs to support physical cameras and the capability could be checked via CameraInfo.isLogicalMultiCameraSupported. Each physical cameras can have its own UseCases but needs to have the same LifecycleOwner, otherwise IllegalArgumentException will be thrown.

If we want to open one physical camera, for example ultra wide, we just need to set physical camera id in CameraSelector and bind to lifecycle. All CameraX features will work normally when only a single physical camera is used.

If we want to open multiple physical cameras, we need to have multiple CameraSelectors, each in one SingleCameraConfig and set physical camera id, then bind to lifecycle with the SingleCameraConfigs. Internally each physical camera id will be set on UseCase, for example, Preview and call android.hardware.camera2.params.OutputConfiguration.setPhysicalCameraId.

Currently only two physical cameras for the same logical camera id are allowed and the device needs to support physical cameras by checking CameraInfo.isLogicalMultiCameraSupported. In addition, there is no guarantee or API to query whether the device supports multiple physical camera opening or not. Internally the library checks android.hardware.camera2.CameraDevice.isSessionConfigurationSupported, if the device does not support the multiple physical camera configuration, IllegalArgumentException will be thrown.

@MainThread
fun bindToLifecycle(
    lifecycleOwner: LifecycleOwner,
    cameraSelector: CameraSelector,
    useCaseGroup: UseCaseGroup
): Camera

Binds a UseCaseGroup to a LifecycleOwner.

Similar to bindToLifecycle, with the addition that the bound collection of UseCase share parameters defined by UseCaseGroup such as consistent camera sensor rect across all UseCases.

If one UseCase is in multiple UseCaseGroups, it will be linked to the UseCaseGroup in the latest bindToLifecycle call.

@MainThread
fun bindToLifecycle(
    lifecycleOwner: LifecycleOwner,
    cameraSelector: CameraSelector,
    vararg useCases: UseCase?
): Camera

Binds the collection of UseCase to a LifecycleOwner.

The state of the lifecycle will determine when the cameras are open, started, stopped and closed. When started, the use cases receive camera data.

Binding to a lifecycleOwner in state currently in Lifecycle.State.STARTED or greater will also initialize and start data capture. If the camera was already running this may cause a new initialization to occur temporarily stopping data from the camera before restarting it.

Multiple use cases can be bound via adding them all to a single bindToLifecycle call, or by using multiple bindToLifecycle calls. Using a single call that includes all the use cases helps to set up a camera session correctly for all uses cases, such as by allowing determination of resolutions depending on all the use cases bound being bound. If the use cases are bound separately, it will find the supported resolution with the priority depending on the binding sequence. If the use cases are bound with a single call, it will find the supported resolution with the priority in sequence of ImageCapture, Preview and then ImageAnalysis. The resolutions that can be supported depends on the camera device hardware level that there are some default guaranteed resolutions listed in android.hardware.camera2.CameraDevice.createCaptureSession.

Currently up to 3 use cases may be bound to a Lifecycle at any time. Exceeding capability of target camera device will throw an IllegalArgumentException.

A UseCase should only be bound to a single lifecycle and camera selector a time. Attempting to bind a use case to a lifecycle when it is already bound to another lifecycle is an error, and the use case binding will not change. Attempting to bind the same use case to multiple camera selectors is also an error and will not change the binding.

If different use cases are bound to different camera selectors that resolve to distinct cameras, but the same lifecycle, only one of the cameras will operate at a time. The non-operating camera will not become active until it is the only camera with use cases bound.

The Camera returned is determined by the given camera selector, plus other internal requirements, possibly from use case configurations. The camera returned from bindToLifecycle may differ from the camera determined solely by a camera selector. If the camera selector can't resolve a valid camera under the requirements, an IllegalArgumentException will be thrown.

Only UseCase bound to latest active Lifecycle can keep alive. UseCase bound to other Lifecycle will be stopped.

open fun getCameraInfo(cameraSelector: CameraSelector): CameraInfo

Returns the CameraInfo instance of the camera resulted from the specified CameraSelector.

The returned CameraInfo corresponds to the camera that will be bound when calling bindToLifecycle with the specified CameraSelector.

open fun hasCamera(cameraSelector: CameraSelector): Boolean

Checks whether this provider supports at least one camera that meets the requirements from a CameraSelector.

If this method returns true, then the camera selector can be used to bind use cases and retrieve a Camera instance.

fun isBound(useCase: UseCase): Boolean

Returns true if this UseCase is bound to a lifecycle or included in a bound SessionConfig. Otherwise returns false.

After binding a use case, use cases remain bound until the lifecycle reaches a Lifecycle.State.DESTROYED state or if is unbound by calls to unbind or unbindAll.

@VisibleForTesting
fun shutdownAsync(): ListenableFuture<Void>

Allows shutting down this ProcessCameraProvider instance so a new instance can be retrieved by getInstance.

Once shutdownAsync is invoked, a new instance can be retrieved with getInstance.

This method should be used for testing purposes only. Along with configureInstance, this allows the process camera provider to be used in test suites which may need to initialize CameraX in different ways in between tests.

@MainThread
fun unbind(vararg useCases: UseCase?): Unit

Unbinds all specified use cases from the lifecycle provider.

This will initiate a close of every open camera which has zero UseCase associated with it at the end of this call.

If a use case in the argument list is not bound, then it is simply ignored.

After unbinding a UseCase, the UseCase can be bound to another Lifecycle however listeners and settings should be reset by the application.

@MainThread
fun unbindAll(): Unit

Unbinds all use cases from the lifecycle provider and removes them from CameraX.

This will initiate a close of every currently open camera.

open val availableCameraInfos: List<CameraInfo>

The CameraInfo instances of the available cameras.

While iterating through all the available CameraInfo, if one of them meets some predefined requirements, a CameraSelector that uniquely identifies its camera can be retrieved using CameraInfo.getCameraSelector, which can then be used to bind use cases to that camera.

open val availableConcurrentCameraInfos: List<List<CameraInfo>>

Returns list of CameraInfo instances of the available concurrent cameras.

The available concurrent cameras include all combinations of cameras which could operate concurrently on the device. Each list maps to one combination of these camera's CameraInfo.

For example, to select a front camera and a back camera and bind to LifecycleOwner with preview UseCase, this function could be used with bindToLifecycle.

import androidx.camera.core.CameraSelector
import androidx.camera.core.ConcurrentCamera.SingleCameraConfig
import androidx.camera.core.Preview
import androidx.camera.core.UseCaseGroup
import androidx.camera.view.PreviewView

var cameraSelectorPrimary: CameraSelector? = null
var cameraSelectorSecondary: CameraSelector? = null
for (cameraInfoList in cameraProvider.availableConcurrentCameraInfos) {
    for (cameraInfo in cameraInfoList) {
        if (cameraInfo.lensFacing == CameraSelector.LENS_FACING_FRONT) {
            cameraSelectorPrimary = cameraInfo.getCameraSelector()
        } else if (cameraInfo.lensFacing == CameraSelector.LENS_FACING_BACK) {
            cameraSelectorSecondary = cameraInfo.getCameraSelector()
        }
    }
}
if (cameraSelectorPrimary == null || cameraSelectorSecondary == null) {
    return
}
val previewFront = Preview.Builder().build()
previewFront.surfaceProvider = frontPreviewView.getSurfaceProvider()
val primary =
    SingleCameraConfig(
        cameraSelectorPrimary,
        UseCaseGroup.Builder().addUseCase(previewFront).build(),
        lifecycleOwner
    )
val previewBack = Preview.Builder().build()
previewBack.surfaceProvider = backPreviewView.getSurfaceProvider()
val secondary =
    SingleCameraConfig(
        cameraSelectorSecondary,
        UseCaseGroup.Builder().addUseCase(previewBack).build(),
        lifecycleOwner
    )
cameraProvider.bindToLifecycle(listOf(primary, secondary))

open val isConcurrentCameraModeOn: Boolean

Returns whether there is a ConcurrentCamera bound.