diff --git a/mediapipe/tasks/ios/vision/gesture_recognizer/BUILD b/mediapipe/tasks/ios/vision/gesture_recognizer/BUILD index d9a76afde..e5b46d344 100644 --- a/mediapipe/tasks/ios/vision/gesture_recognizer/BUILD +++ b/mediapipe/tasks/ios/vision/gesture_recognizer/BUILD @@ -58,7 +58,7 @@ objc_library( "//mediapipe/tasks/ios/core:MPPTaskInfo", "//mediapipe/tasks/ios/vision/core:MPPImage", "//mediapipe/tasks/ios/vision/core:MPPVisionPacketCreator", - "//mediapipe/tasks/ios/vision/core:MPPVisionTaskRunner", + "//mediapipe/tasks/ios/vision/core:MPPVisionTaskRunnerRefactored", "//mediapipe/tasks/ios/vision/gesture_recognizer/utils:MPPGestureRecognizerOptionsHelpers", "//mediapipe/tasks/ios/vision/gesture_recognizer/utils:MPPGestureRecognizerResultHelpers", ], diff --git a/mediapipe/tasks/ios/vision/gesture_recognizer/sources/MPPGestureRecognizer.mm b/mediapipe/tasks/ios/vision/gesture_recognizer/sources/MPPGestureRecognizer.mm index b722163b2..3853f6b02 100644 --- a/mediapipe/tasks/ios/vision/gesture_recognizer/sources/MPPGestureRecognizer.mm +++ b/mediapipe/tasks/ios/vision/gesture_recognizer/sources/MPPGestureRecognizer.mm @@ -18,7 +18,7 @@ #import "mediapipe/tasks/ios/common/utils/sources/NSString+Helpers.h" #import "mediapipe/tasks/ios/core/sources/MPPTaskInfo.h" #import "mediapipe/tasks/ios/vision/core/sources/MPPVisionPacketCreator.h" -#import "mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.h" +#import "mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunnerRefactored.h" #import "mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerOptions+Helpers.h" #import "mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.h" @@ -54,6 +54,19 @@ static NSString *const kTaskName = @"gestureRecognizer"; } \ } +#define GestureRecognizerResultWithOutputPacketMap(outputPacketMap) \ + { \ + [MPPGestureRecognizerResult \ + gestureRecognizerResultWithHandGesturesPacket:outputPacketMap[kHandGesturesOutStreamName \ + .cppString] \ + handednessPacket:outputPacketMap[kHandednessOutStreamName \ + .cppString] \ + handLandmarksPacket:outputPacketMap[kLandmarksOutStreamName \ + .cppString] \ + worldLandmarksPacket:outputPacketMap[kWorldLandmarksOutStreamName \ + .cppString]] \ + } + @interface MPPGestureRecognizer () { /** iOS Vision Task Runner */ MPPVisionTaskRunner *_visionTaskRunner; @@ -65,56 +78,6 @@ static NSString *const kTaskName = @"gestureRecognizer"; @implementation MPPGestureRecognizer -- (nullable MPPGestureRecognizerResult *)gestureRecognizerResultWithOutputPacketMap: - (PacketMap &)outputPacketMap { - return [MPPGestureRecognizerResult - gestureRecognizerResultWithHandGesturesPacket:outputPacketMap[kHandGesturesOutStreamName - .cppString] - handednessPacket:outputPacketMap[kHandednessOutStreamName - .cppString] - handLandmarksPacket:outputPacketMap[kLandmarksOutStreamName - .cppString] - worldLandmarksPacket:outputPacketMap[kWorldLandmarksOutStreamName - .cppString]]; -} - -- (void)processLiveStreamResult:(absl::StatusOr)liveStreamResult { - if (![self.gestureRecognizerLiveStreamDelegate - respondsToSelector:@selector(gestureRecognizer: - didFinishRecognitionWithResult:timestampInMilliseconds:error:)]) { - return; - } - - NSError *callbackError = nil; - if (![MPPCommonUtils checkCppError:liveStreamResult.status() toError:&callbackError]) { - dispatch_async(_callbackQueue, ^{ - [self.gestureRecognizerLiveStreamDelegate gestureRecognizer:self - didFinishRecognitionWithResult:nil - timestampInMilliseconds:Timestamp::Unset().Value() - error:callbackError]; - }); - return; - } - - PacketMap &outputPacketMap = liveStreamResult.value(); - if (outputPacketMap[kImageOutStreamName.cppString].IsEmpty()) { - return; - } - - MPPGestureRecognizerResult *result = - [self gestureRecognizerResultWithOutputPacketMap:outputPacketMap]; - - NSInteger timeStampInMilliseconds = - outputPacketMap[kImageOutStreamName.cppString].Timestamp().Value() / - kMicroSecondsPerMilliSecond; - dispatch_async(_callbackQueue, ^{ - [self.gestureRecognizerLiveStreamDelegate gestureRecognizer:self - didFinishRecognitionWithResult:result - timestampInMilliseconds:timeStampInMilliseconds - error:callbackError]; - }); -} - - (instancetype)initWithOptions:(MPPGestureRecognizerOptions *)options error:(NSError **)error { self = [super init]; if (self) { @@ -161,11 +124,13 @@ static NSString *const kTaskName = @"gestureRecognizer"; }; } - _visionTaskRunner = - [[MPPVisionTaskRunner alloc] initWithCalculatorGraphConfig:[taskInfo generateGraphConfig] - runningMode:options.runningMode - packetsCallback:std::move(packetsCallback) - error:error]; + _visionTaskRunner = [[MPPVisionTaskRunner alloc] initWithTaskInfo:taskInfo + runningMode:options.runningMode + roiAllowed:NO + packetsCallback:std::move(packetsCallback) + imageInputStreamName:kImageInStreamName + normRectInputStreamName:kNormRectInStreamName + error:error]; if (!_visionTaskRunner) { return nil; } @@ -181,93 +146,76 @@ static NSString *const kTaskName = @"gestureRecognizer"; return [self initWithOptions:options error:error]; } -- (nullable MPPGestureRecognizerResult *)gestureRecognizerResultWithOptionalOutputPacketMap: - (std::optional &)outputPacketMap { - if (!outputPacketMap.has_value()) { - return nil; - } - MPPGestureRecognizerResult *result = - [self gestureRecognizerResultWithOutputPacketMap:outputPacketMap.value()]; - return result; -} - - (nullable MPPGestureRecognizerResult *)recognizeImage:(MPPImage *)image error:(NSError **)error { - std::optional rect = - [_visionTaskRunner normalizedRectWithImageOrientation:image.orientation - imageSize:CGSizeMake(image.width, image.height) - error:error]; - if (!rect.has_value()) { - return nil; - } + std::optional outputPacketMap = [_visionTaskRunner processImage:image error:error]; - Packet imagePacket = [MPPVisionPacketCreator createPacketWithMPPImage:image error:error]; - if (imagePacket.IsEmpty()) { - return nil; - } - - Packet normalizedRectPacket = - [MPPVisionPacketCreator createPacketWithNormalizedRect:rect.value()]; - - PacketMap inputPacketMap = InputPacketMap(imagePacket, normalizedRectPacket); - - std::optional outputPacketMap = [_visionTaskRunner processImagePacketMap:inputPacketMap - error:error]; - return [self gestureRecognizerResultWithOptionalOutputPacketMap:outputPacketMap]; -} - -- (std::optional)inputPacketMapWithMPPImage:(MPPImage *)image - timestampInMilliseconds:(NSInteger)timestampInMilliseconds - error:(NSError **)error { - std::optional rect = - [_visionTaskRunner normalizedRectWithImageOrientation:image.orientation - imageSize:CGSizeMake(image.width, image.height) - error:error]; - if (!rect.has_value()) { - return std::nullopt; - } - - Packet imagePacket = [MPPVisionPacketCreator createPacketWithMPPImage:image - timestampInMilliseconds:timestampInMilliseconds - error:error]; - if (imagePacket.IsEmpty()) { - return std::nullopt; - } - - Packet normalizedRectPacket = - [MPPVisionPacketCreator createPacketWithNormalizedRect:rect.value() - timestampInMilliseconds:timestampInMilliseconds]; - - PacketMap inputPacketMap = InputPacketMap(imagePacket, normalizedRectPacket); - return inputPacketMap; + return [MPPGestureRecognizer gestureRecognizerResultWithOptionalOutputPacketMap:outputPacketMap]; } - (nullable MPPGestureRecognizerResult *)recognizeVideoFrame:(MPPImage *)image timestampInMilliseconds:(NSInteger)timestampInMilliseconds error:(NSError **)error { - std::optional inputPacketMap = [self inputPacketMapWithMPPImage:image - timestampInMilliseconds:timestampInMilliseconds - error:error]; - if (!inputPacketMap.has_value()) { - return nil; - } - std::optional outputPacketMap = - [_visionTaskRunner processVideoFramePacketMap:inputPacketMap.value() error:error]; + [_visionTaskRunner processVideoFrame:image + timestampInMilliseconds:timestampInMilliseconds + error:error]; - return [self gestureRecognizerResultWithOptionalOutputPacketMap:outputPacketMap]; + return [MPPGestureRecognizer gestureRecognizerResultWithOptionalOutputPacketMap:outputPacketMap]; } - (BOOL)recognizeAsyncImage:(MPPImage *)image timestampInMilliseconds:(NSInteger)timestampInMilliseconds error:(NSError **)error { - std::optional inputPacketMap = [self inputPacketMapWithMPPImage:image - timestampInMilliseconds:timestampInMilliseconds - error:error]; - if (!inputPacketMap.has_value()) { - return NO; + return [_visionTaskRunner processLiveStreamImage:image + timestampInMilliseconds:timestampInMilliseconds + error:error]; +} + +#pragma mark - Private + +- (void)processLiveStreamResult:(absl::StatusOr)liveStreamResult { + if (![self.gestureRecognizerLiveStreamDelegate + respondsToSelector:@selector(gestureRecognizer: + didFinishRecognitionWithResult:timestampInMilliseconds:error:)]) { + return; } - return [_visionTaskRunner processLiveStreamPacketMap:inputPacketMap.value() error:error]; + NSError *callbackError = nil; + if (![MPPCommonUtils checkCppError:liveStreamResult.status() toError:&callbackError]) { + dispatch_async(_callbackQueue, ^{ + [self.gestureRecognizerLiveStreamDelegate gestureRecognizer:self + didFinishRecognitionWithResult:nil + timestampInMilliseconds:Timestamp::Unset().Value() + error:callbackError]; + }); + return; + } + + PacketMap &outputPacketMap = liveStreamResult.value(); + if (outputPacketMap[kImageOutStreamName.cppString].IsEmpty()) { + return; + } + + MPPGestureRecognizerResult *result = GestureRecognizerResultWithOutputPacketMap(outputPacketMap); + + NSInteger timeStampInMilliseconds = + outputPacketMap[kImageOutStreamName.cppString].Timestamp().Value() / + kMicroSecondsPerMilliSecond; + dispatch_async(_callbackQueue, ^{ + [self.gestureRecognizerLiveStreamDelegate gestureRecognizer:self + didFinishRecognitionWithResult:result + timestampInMilliseconds:timeStampInMilliseconds + error:callbackError]; + }); +} + ++ (nullable MPPGestureRecognizerResult *)gestureRecognizerResultWithOptionalOutputPacketMap: + (std::optional &)outputPacketMap { + if (!outputPacketMap.has_value()) { + return nil; + } + + return GestureRecognizerResultWithOutputPacketMap(outputPacketMap.value()); } @end