ImageDescriber

  • ImageDescriber provides an interface for performing both standard and streaming image description inferences.

  • It includes methods for checking feature status, downloading required assets, preparing the inference engine, and releasing resources.

  • Image description is performed by calling runInference with an ImageDescriptionRequest, which returns a ListenableFuture.

  • Inference requests can be cancelled and may fail due to invalid input or exceeding usage quotas.
public class ImageDescriber extends Object

Provides an interface for performing image description.

It supports both standard and streaming inferences, as well as methods for preparing and cleaning up model resources.

Typical usage:


 ImageDescriptionRequest request = ImageDescriptionRequest.builder(bitmap).build();

 ListenableFuture future = imageDescriber.runInference(request);
 Futures.addCallback(future, new FutureCallback<>() {
     @Override
     public void onSuccess(ImageDescriptionResult result) {
         Log.d(TAG, result.getDescription());
     }

     @Override
     public void onFailure(Throwable t) {
         Log.e(TAG, "Failed to describe", t);
     }
 }, executor);

Public Method Summary

ListenableFuture<Integer>
checkFeatureStatus()
Checks the current availability status of the image description feature.
void
close()
Releases resources associated with the image description engine.
ListenableFuture<Void>
downloadFeature(DownloadCallback callback)
Downloads the required model assets for the image description feature if they are not already available.
ListenableFuture<String>
getBaseModelName()
Returns the name of the base model used by this image describer instance.
ListenableFuture<Void>
prepareInferenceEngine()
Prepares the inference engine for use by loading necessary models and initializing runtime components.
ListenableFuture<ImageDescriptionResult>
runInference(ImageDescriptionRequest request, StreamingCallback streamingCallback)
Performs streaming image description inference on the provided image request.
ListenableFuture<ImageDescriptionResult>
runInference(ImageDescriptionRequest request)
Performs asynchronous image description on the provided image request.

Inherited Method Summary

Public Methods

public ListenableFuture<Integer> checkFeatureStatus ()

Checks the current availability status of the image description feature.

Returns

public void close ()

Releases resources associated with the image description engine.

This should be called when the image describer is no longer needed. Can be safely called multiple times.

public ListenableFuture<Void> downloadFeature (DownloadCallback callback)

Downloads the required model assets for the image description feature if they are not already available.

Use this method to proactively download models before inference. The provided DownloadCallback is invoked to report progress and completion status.

Parameters
callback a non-null DownloadCallback for receiving download updates.
Returns
  • a ListenableFuture that completes when the download finishes or fails. It completes immediately if already downloaded.

public ListenableFuture<String> getBaseModelName ()

Returns the name of the base model used by this image describer instance.

The model name may be used for logging, debugging, or feature gating purposes.

Returns

public ListenableFuture<Void> prepareInferenceEngine ()

Prepares the inference engine for use by loading necessary models and initializing runtime components.

If the models haven't been downloaded yet, it will trigger the download and wait for it to complete first.

While calling this method is optional, we recommend invoking it well before the first inference call to reduce the latency of the initial inference.

Returns

public ListenableFuture<ImageDescriptionResult> runInference (ImageDescriptionRequest request, StreamingCallback streamingCallback)

Performs streaming image description inference on the provided image request.

Partial results are delivered incrementally through the provided StreamingCallback. This is useful to build more responsive UI. Streaming can be interrupted by a GenAiException. In that case, consider removing any already streamed partial output from the UI.

The returned ListenableFuture is cancellable. If the inference is no longer needed (e.g., the user navigates away), calling future.cancel(true) will attempt to interrupt the process and free up resources.

Note that inference requests may fail under certain conditions such as:

Appropriate GenAiException handling should be implemented when attaching the callback to the future.
Parameters
request a non-null ImageDescriptionRequest containing the input image.
streamingCallback a non-null StreamingCallback for receiving streamed results.
Returns

public ListenableFuture<ImageDescriptionResult> runInference (ImageDescriptionRequest request)

Performs asynchronous image description on the provided image request.

This is the standard, non-streaming version of inference. The full description is returned once the model completes processing.

This method is non-blocking. To handle the result, callers should attach a listener to the returned ListenableFuture using Futures.addCallback(ListenableFuture, FutureCallback , Executor) or similar methods. The inference runs on background threads and may complete at a later time depending on model availability and input size.

The returned ListenableFuture is cancellable. If the inference is no longer needed (e.g., the user navigates away), calling future.cancel(true) will attempt to interrupt the process and free up resources.

Note that inference requests may fail under certain conditions such as:

Appropriate GenAiException handling should be implemented when attaching the callback to the future.
Parameters
request a non-null ImageDescriptionRequest containing input image.
Returns