diff --git a/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.h b/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.h index 9a11d1e2..9ad95847 100644 --- a/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.h +++ b/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.h @@ -19,8 +19,30 @@ NS_ASSUME_NONNULL_BEGIN @interface MPPCategory (Helpers) +/** + * Creates an `MPPCategory` with the given MediaPipe `Classification` proto. + * + * @param classificationProto A MediaPipe `Classification` proto. + * @return An `MPPCategory` object that with the given MediaPipe `Classification` proto. + */ + (MPPCategory *)categoryWithProto:(const ::mediapipe::Classification &)classificationProto; +/** + * Creates an `MPPCategory` with the given MediaPipe `Classification` proto and the given category + * index. The resulting `MPPCategory` is created with the given category index instead of the + * category index specified in the `Classification` proto. This method is useful for tasks like + * gesture recognizer which always returns a default index for the recognized gestures. + * + * @param classificationProto A MediaPipe `Classification` proto. + * @param index The index to be used for creating the `MPPCategory` instead of the category index + * specified in the `Classification` proto. + * + * @return An `MPPGestureRecognizerResult` object that contains the hand gesture recognition + * results. + */ ++ (MPPCategory *)categoryWithProto:(const ::mediapipe::Classification &)classificationProto + index:(NSInteger)index; + @end NS_ASSUME_NONNULL_END diff --git a/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.mm b/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.mm index 12cfa562..542b8c41 100644 --- a/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.mm +++ b/mediapipe/tasks/ios/components/containers/utils/sources/MPPCategory+Helpers.mm @@ -21,7 +21,8 @@ using ClassificationProto = ::mediapipe::Classification; @implementation MPPCategory (Helpers) -+ (MPPCategory *)categoryWithProto:(const ClassificationProto &)classificationProto { ++ (MPPCategory *)categoryWithProto:(const ClassificationProto &)classificationProto + index:(NSInteger)index { NSString *categoryName; NSString *displayName; @@ -33,10 +34,14 @@ using ClassificationProto = ::mediapipe::Classification; displayName = [NSString stringWithCppString:classificationProto.display_name()]; } - return [[MPPCategory alloc] initWithIndex:classificationProto.index() + return [[MPPCategory alloc] initWithIndex:index score:classificationProto.score() categoryName:categoryName displayName:displayName]; } ++ (MPPCategory *)categoryWithProto:(const ClassificationProto &)classificationProto { + return [MPPCategory categoryWithProto:classificationProto index:classificationProto.index()]; +} + @end diff --git a/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.h b/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.h index 318b2405..a7930244 100644 --- a/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.h +++ b/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.h @@ -58,38 +58,57 @@ NS_ASSUME_NONNULL_BEGIN error:(NSError **)error NS_DESIGNATED_INITIALIZER; /** - * Creates a `NormalizedRect` from a region of interest and an image orientation, performing - * sanity checks on-the-fly. - * If the input region of interest equals `CGRectZero`, returns a default `NormalizedRect` covering - * the whole image with rotation set according `imageOrientation`. If `ROIAllowed` is NO, an error - * will be returned if the input region of interest is not equal to `CGRectZero`. Mirrored - * orientations (`UIImageOrientationUpMirrored`,`UIImageOrientationDownMirrored`, + * Creates a `NormalizedRect` from image orientation for a task which does not support roi, + * performing sanity checks on-the-fly. Mirrored orientations + * (`UIImageOrientationUpMirrored`,`UIImageOrientationDownMirrored`, * `UIImageOrientationLeftMirrored`,`UIImageOrientationRightMirrored`) are not supported. An error * will be returned if `imageOrientation` is equal to any one of them. * - * @param roi A `CGRect` specifying the region of interest. If the input region of interest equals - * `CGRectZero`, the returned `NormalizedRect` covers the whole image. Make sure that `roi` equals - * `CGRectZero` if `ROIAllowed` is NO. Otherwise, an error will be returned. - * @param imageSize A `CGSize` specifying the size of the image within which normalized rect is - * calculated. * @param imageOrientation A `UIImageOrientation` indicating the rotation to be applied to the * image. The resulting `NormalizedRect` will convert the `imageOrientation` to degrees clockwise. * Mirrored orientations (`UIImageOrientationUpMirrored`, `UIImageOrientationDownMirrored`, * `UIImageOrientationLeftMirrored`, `UIImageOrientationRightMirrored`) are not supported. An error * will be returned if `imageOrientation` is equal to any one of them. - * @param ROIAllowed Indicates if the `roi` field is allowed to be a value other than `CGRectZero`. + * @param imageSize A `CGSize` specifying the size of the image within which normalized rect is + * calculated. + * @param error Pointer to the memory location where errors if any should be saved. If @c NULL, no + * error will be saved. + * + * @return An optional `NormalizedRect` from the given region of interest and image orientation. + */ +- (std::optional)normalizedRectWithImageOrientation: + (UIImageOrientation)imageOrientation + imageSize:(CGSize)imageSize + error:(NSError **)error; + +/** + * Creates a `NormalizedRect` from roi and image orientation for a task which supports roi, + * performing sanity checks on-the-fly. If the input region of interest equals `CGRectZero`, returns + * a default `NormalizedRect` covering the whole image with rotation set according + * `imageOrientation`. Mirrored orientations + * (`UIImageOrientationUpMirrored`,`UIImageOrientationDownMirrored`, + * `UIImageOrientationLeftMirrored`,`UIImageOrientationRightMirrored`) are not supported. An error + * will be returned if `imageOrientation` is equal to any one of them. + * + * @param roi A `CGRect` specifying the region of interest. If the input region of interest equals + * `CGRectZero`, the returned `NormalizedRect` covers the whole image. + * @param imageOrientation A `UIImageOrientation` indicating the rotation to be applied to the + * image. The resulting `NormalizedRect` will convert the `imageOrientation` to degrees clockwise. + * Mirrored orientations (`UIImageOrientationUpMirrored`, `UIImageOrientationDownMirrored`, + * `UIImageOrientationLeftMirrored`, `UIImageOrientationRightMirrored`) are not supported. An error + * will be returned if `imageOrientation` is equal to any one of them. + * @param imageSize A `CGSize` specifying the size of the image within which normalized rect is + * calculated. * @param error Pointer to the memory location where errors if any should be saved. If @c NULL, no * error will be saved. * * @return An optional `NormalizedRect` from the given region of interest and image orientation. */ - (std::optional) - normalizedRectFromRegionOfInterest:(CGRect)roi - imageSize:(CGSize)imageSize + normalizedRectWithRegionOfInterest:(CGRect)roi imageOrientation:(UIImageOrientation)imageOrientation - ROIAllowed:(BOOL)ROIAllowed + imageSize:(CGSize)imageSize error:(NSError **)error; - /** * A synchronous method to invoke the C++ task runner to process single image inputs. The call * blocks the current thread until a failure status or a successful result is returned. diff --git a/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.mm b/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.mm index 0089e516..c1b5d058 100644 --- a/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.mm +++ b/mediapipe/tasks/ios/vision/core/sources/MPPVisionTaskRunner.mm @@ -91,7 +91,30 @@ static NSString *const kTaskPrefix = @"com.mediapipe.tasks.vision"; return self; } -- (std::optional)normalizedRectFromRegionOfInterest:(CGRect)roi +- (std::optional)normalizedRectWithRegionOfInterest:(CGRect)roi + imageOrientation: + (UIImageOrientation)imageOrientation + imageSize:(CGSize)imageSize + error:(NSError **)error { + return [self normalizedRectWithRegionOfInterest:roi + imageSize:imageSize + imageOrientation:imageOrientation + ROIAllowed:YES + error:error]; +} + +- (std::optional)normalizedRectWithImageOrientation: + (UIImageOrientation)imageOrientation + imageSize:(CGSize)imageSize + error:(NSError **)error { + return [self normalizedRectWithRegionOfInterest:CGRectZero + imageSize:imageSize + imageOrientation:imageOrientation + ROIAllowed:NO + error:error]; +} + +- (std::optional)normalizedRectWithRegionOfInterest:(CGRect)roi imageSize:(CGSize)imageSize imageOrientation: (UIImageOrientation)imageOrientation diff --git a/mediapipe/tasks/ios/vision/face_detector/sources/MPPFaceDetector.mm b/mediapipe/tasks/ios/vision/face_detector/sources/MPPFaceDetector.mm index ceb5c957..7cb525fb 100644 --- a/mediapipe/tasks/ios/vision/face_detector/sources/MPPFaceDetector.mm +++ b/mediapipe/tasks/ios/vision/face_detector/sources/MPPFaceDetector.mm @@ -128,10 +128,8 @@ static NSString *const kTaskName = @"faceDetector"; timestampInMilliseconds:(NSInteger)timestampInMilliseconds error:(NSError **)error { std::optional rect = - [_visionTaskRunner normalizedRectFromRegionOfInterest:CGRectZero + [_visionTaskRunner normalizedRectWithImageOrientation:image.orientation imageSize:CGSizeMake(image.width, image.height) - imageOrientation:image.orientation - ROIAllowed:NO error:error]; if (!rect.has_value()) { return std::nullopt; @@ -154,10 +152,8 @@ static NSString *const kTaskName = @"faceDetector"; - (nullable MPPFaceDetectorResult *)detectInImage:(MPPImage *)image error:(NSError **)error { std::optional rect = - [_visionTaskRunner normalizedRectFromRegionOfInterest:CGRectZero + [_visionTaskRunner normalizedRectWithImageOrientation:image.orientation imageSize:CGSizeMake(image.width, image.height) - imageOrientation:image.orientation - ROIAllowed:NO error:error]; if (!rect.has_value()) { return nil; diff --git a/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.h b/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.h index 649c11c8..6b0f8bf8 100644 --- a/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.h +++ b/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.h @@ -14,6 +14,8 @@ #import "mediapipe/tasks/ios/vision/gesture_recognizer/sources/MPPGestureRecognizerResult.h" +#include "mediapipe/framework/formats/classification.pb.h" +#include "mediapipe/framework/formats/landmark.pb.h" #include "mediapipe/framework/packet.h" NS_ASSUME_NONNULL_BEGIN @@ -23,14 +25,14 @@ static const int kMicroSecondsPerMilliSecond = 1000; @interface MPPGestureRecognizerResult (Helpers) /** - * Creates an `MPPGestureRecognizerResult` from hand gestures, handedness, hand landmarks and world + * Creates an `MPPGestureRecognizerResult` from hand gestures, handedness, hand landmarks and world * landmarks packets. * - * @param handGesturesPacket a MediaPipe packet wrapping a`std::vector`. - * @param handednessPacket a MediaPipe packet wrapping a`std::vector`. - * @param handLandmarksPacket a MediaPipe packet wrapping + * @param handGesturesPacket A MediaPipe packet wrapping a`std::vector`. + * @param handednessPacket A MediaPipe packet wrapping a`std::vector`. + * @param handLandmarksPacket A MediaPipe packet wrapping * a`std::vector`. - * @param handLandmarksPacket a MediaPipe packet wrapping a`std::vector`. + * @param worldLandmarksPacket A MediaPipe packet wrapping a`std::vector`. * * @return An `MPPGestureRecognizerResult` object that contains the hand gesture recognition * results. diff --git a/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.mm b/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.mm index 70773a94..8eed2a92 100644 --- a/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.mm +++ b/mediapipe/tasks/ios/vision/gesture_recognizer/utils/sources/MPPGestureRecognizerResult+Helpers.mm @@ -21,6 +21,8 @@ #include "mediapipe/framework/formats/landmark.pb.h" #include "mediapipe/framework/packet.h" +static const NSInteger kDefaultGestureIndex = -1; + namespace { using ClassificationListProto = ::mediapipe::ClassificationList; using LandmarkListProto = ::mediapipe::LandmarkList; @@ -30,6 +32,15 @@ using ::mediapipe::Packet; @implementation MPPGestureRecognizerResult (Helpers) ++ (MPPGestureRecognizerResult *)emptyGestureRecognizerResultWithTimestampInMilliseconds: + (NSInteger)timestampInMilliseconds { + return [[MPPGestureRecognizerResult alloc] initWithGestures:@[] + handedness:@[] + landmarks:@[] + worldLandmarks:@[] + timestampInMilliseconds:timestampInMilliseconds]; +} + + (MPPGestureRecognizerResult *) gestureRecognizerResultWithHandGesturesPacket:(const Packet &)handGesturesPacket handednessPacket:(const Packet &)handednessPacket @@ -39,18 +50,16 @@ using ::mediapipe::Packet; (NSInteger)(handGesturesPacket.Timestamp().Value() / kMicroSecondsPerMilliSecond); if (handGesturesPacket.IsEmpty()) { - return [[MPPGestureRecognizerResult alloc] initWithGestures:@[] - handedness:@[] - landmarks:@[] - worldLandmarks:@[] - timestampInMilliseconds:timestampInMilliseconds]; + return [MPPGestureRecognizerResult + emptyGestureRecognizerResultWithTimestampInMilliseconds:timestampInMilliseconds]; } if (!handGesturesPacket.ValidateAsType>().ok() || !handednessPacket.ValidateAsType>().ok() || !handLandmarksPacket.ValidateAsType>().ok() || !worldLandmarksPacket.ValidateAsType>().ok()) { - return nil; + return [MPPGestureRecognizerResult + emptyGestureRecognizerResultWithTimestampInMilliseconds:timestampInMilliseconds]; } const std::vector &handGesturesClassificationListProtos = @@ -62,7 +71,8 @@ using ::mediapipe::Packet; NSMutableArray *gestures = [NSMutableArray arrayWithCapacity:(NSUInteger)classificationListProto.classification().size()]; for (const auto &classificationProto : classificationListProto.classification()) { - MPPCategory *category = [MPPCategory categoryWithProto:classificationProto]; + MPPCategory *category = [MPPCategory categoryWithProto:classificationProto + index:kDefaultGestureIndex]; [gestures addObject:category]; } [multiHandGestures addObject:gestures]; diff --git a/mediapipe/tasks/ios/vision/image_classifier/sources/MPPImageClassifier.mm b/mediapipe/tasks/ios/vision/image_classifier/sources/MPPImageClassifier.mm index 3ad8d0de..10f6fc26 100644 --- a/mediapipe/tasks/ios/vision/image_classifier/sources/MPPImageClassifier.mm +++ b/mediapipe/tasks/ios/vision/image_classifier/sources/MPPImageClassifier.mm @@ -166,10 +166,9 @@ static const int kMicroSecondsPerMilliSecond = 1000; regionOfInterest:(CGRect)roi error:(NSError **)error { std::optional rect = - [_visionTaskRunner normalizedRectFromRegionOfInterest:roi - imageSize:CGSizeMake(image.width, image.height) + [_visionTaskRunner normalizedRectWithRegionOfInterest:roi imageOrientation:image.orientation - ROIAllowed:YES + imageSize:CGSizeMake(image.width, image.height) error:error]; if (!rect.has_value()) { return nil; @@ -196,15 +195,18 @@ static const int kMicroSecondsPerMilliSecond = 1000; outputPacketMap.value()[kClassificationsStreamName.cppString]]; } +- (nullable MPPImageClassifierResult *)classifyImage:(MPPImage *)image error:(NSError **)error { + return [self classifyImage:image regionOfInterest:CGRectZero error:error]; +} + - (std::optional)inputPacketMapWithMPPImage:(MPPImage *)image timestampInMilliseconds:(NSInteger)timestampInMilliseconds regionOfInterest:(CGRect)roi error:(NSError **)error { std::optional rect = - [_visionTaskRunner normalizedRectFromRegionOfInterest:roi - imageSize:CGSizeMake(image.width, image.height) + [_visionTaskRunner normalizedRectWithRegionOfInterest:roi imageOrientation:image.orientation - ROIAllowed:YES + imageSize:CGSizeMake(image.width, image.height) error:error]; if (!rect.has_value()) { return std::nullopt; @@ -225,10 +227,6 @@ static const int kMicroSecondsPerMilliSecond = 1000; return inputPacketMap; } -- (nullable MPPImageClassifierResult *)classifyImage:(MPPImage *)image error:(NSError **)error { - return [self classifyImage:image regionOfInterest:CGRectZero error:error]; -} - - (nullable MPPImageClassifierResult *)classifyVideoFrame:(MPPImage *)image timestampInMilliseconds:(NSInteger)timestampInMilliseconds regionOfInterest:(CGRect)roi diff --git a/mediapipe/tasks/ios/vision/object_detector/sources/MPPObjectDetector.mm b/mediapipe/tasks/ios/vision/object_detector/sources/MPPObjectDetector.mm index 27b196d7..b7924996 100644 --- a/mediapipe/tasks/ios/vision/object_detector/sources/MPPObjectDetector.mm +++ b/mediapipe/tasks/ios/vision/object_detector/sources/MPPObjectDetector.mm @@ -160,10 +160,8 @@ static NSString *const kTaskName = @"objectDetector"; timestampInMilliseconds:(NSInteger)timestampInMilliseconds error:(NSError **)error { std::optional rect = - [_visionTaskRunner normalizedRectFromRegionOfInterest:CGRectZero + [_visionTaskRunner normalizedRectWithImageOrientation:image.orientation imageSize:CGSizeMake(image.width, image.height) - imageOrientation:image.orientation - ROIAllowed:NO error:error]; if (!rect.has_value()) { return std::nullopt; @@ -188,10 +186,8 @@ static NSString *const kTaskName = @"objectDetector"; regionOfInterest:(CGRect)roi error:(NSError **)error { std::optional rect = - [_visionTaskRunner normalizedRectFromRegionOfInterest:roi + [_visionTaskRunner normalizedRectWithImageOrientation:image.orientation imageSize:CGSizeMake(image.width, image.height) - imageOrientation:image.orientation - ROIAllowed:NO error:error]; if (!rect.has_value()) { return nil;