Wraps image data for on-device machine learning (ODML) usages.
MlImage
is designed to be an immutable image container, which could be shared
cross-platforms, among different Google ODML frameworks(TFLite Support, MLKit).
It is a common abstraction image that could help to chain different frameworks that adapts
MlImage
together.
To construct an MlImage
, use the provided builders:
MlImage
uses reference counting to maintain internal storage. When it is created the
reference count is 1. Developer can call close()
to reduce reference count to release
internal storage earlier, otherwise Java garbage collection will release the storage eventually.
To extract concrete image, first check MlImage.StorageType
and then use the provided
extractors:
MlImage
will support internal conversion(e.g. Bitmap -> ByteBuffer)
and multiple storages.
Nested Classes
@interface | MlImage.ImageFormat | Specifies the image format of an image. | |
class | MlImage.Internal | Advanced API access for MlImage . |
|
@interface | MlImage.StorageType | Specifies the image container type. |
Constants
int | IMAGE_FORMAT_ALPHA | |
int | IMAGE_FORMAT_JPEG | |
int | IMAGE_FORMAT_NV12 | |
int | IMAGE_FORMAT_NV21 | |
int | IMAGE_FORMAT_RGB | |
int | IMAGE_FORMAT_RGBA | |
int | IMAGE_FORMAT_UNKNOWN | |
int | IMAGE_FORMAT_YUV_420_888 | |
int | IMAGE_FORMAT_YV12 | |
int | IMAGE_FORMAT_YV21 | |
int | STORAGE_TYPE_BITMAP | |
int | STORAGE_TYPE_BYTEBUFFER | |
int | STORAGE_TYPE_IMAGE_PROXY | |
int | STORAGE_TYPE_MEDIA_IMAGE |
Public Methods
synchronized void |
close()
Removes a reference that was previously acquired or init.
|
List<ImageProperties> |