cassdlcm/mediapipe
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'master' into move-headers-script

Browse Source
This commit is contained in:
Craig Labenz 2023-09-27 14:10:00 -04:00
commit 40503acdec
37 changed files with 475 additions and 171 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
WORKSPACE
View File

@ -176,6 +176,25 @@ http_archive(
],
)
# 2023-06-05
# This version of Glog is required for Windows support, but currently causes
# crashes on some Android devices.
http_archive(
name = "com_github_glog_glog_windows",
strip_prefix = "glog-3a0d4d22c5ae0b9a2216988411cfa6bf860cc372",
sha256 = "170d08f80210b82d95563f4723a15095eff1aad1863000e8eeb569c96a98fefb",
urls = [
"https://github.com/google/glog/archive/3a0d4d22c5ae0b9a2216988411cfa6bf860cc372.zip",
],
patches = [
"@//third_party:com_github_glog_glog.diff",
"@//third_party:com_github_glog_glog_windows_patch.diff",
],
patch_args = [
"-p1",
],
)
# easyexif
http_archive(
name = "easyexif",

1
mediapipe/calculators/tensor/image_to_tensor_converter_gl_texture.cc
View File

@ -304,6 +304,7 @@ class GlProcessor : public ImageToTensorConverter {
glBindTexture(GL_TEXTURE_2D, 0);
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, 0);
glFlush();
return absl::OkStatus();
}

1
mediapipe/calculators/tensor/tensor_converter_calculator.cc
View File

@ -406,6 +406,7 @@ absl::Status TensorConverterCalculator::ProcessGPU(CalculatorContext* cc) {
glActiveTexture(GL_TEXTURE1);
glBindTexture(GL_TEXTURE_2D, 0);
#endif // MEDIAPIPE_OPENGL_ES_VERSION >= MEDIAPIPE_OPENGL_ES_31
glFlush();
src.Release();
return absl::OkStatus();
}));

21
mediapipe/model_maker/python/text/text_classifier/BUILD
View File

@ -93,6 +93,23 @@ py_test(
deps = [":dataset"],
)
py_library(
name = "model_with_tokenizer",
srcs = ["model_with_tokenizer.py"],
)
py_test(
name = "model_with_tokenizer_test",
srcs = ["model_with_tokenizer_test.py"],
tags = ["requires-net:external"],
deps = [
":bert_tokenizer",
":model_spec",
":model_with_tokenizer",
"//mediapipe/model_maker/python/core/utils:hub_loader",
],
)
py_library(
name = "bert_tokenizer",
srcs = ["bert_tokenizer.py"],
@ -145,10 +162,12 @@ py_library(
name = "text_classifier",
srcs = ["text_classifier.py"],
deps = [
":bert_tokenizer",
":dataset",
":hyperparameters",
":model_options",
":model_spec",
":model_with_tokenizer",
":preprocessor",
":text_classifier_options",
"//mediapipe/model_maker/python/core/data:dataset",
@ -165,7 +184,7 @@ py_library(
py_test(
name = "text_classifier_test",
size = "large",
size = "enormous",
srcs = ["text_classifier_test.py"],
data = [
"//mediapipe/model_maker/python/text/text_classifier/testdata",

41
mediapipe/model_maker/python/text/text_classifier/bert_tokenizer.py
View File

@ -56,6 +56,15 @@ class BertFullTokenizer(BertTokenizer):
self._seq_len = seq_len
def process(self, input_tensor: tf.Tensor) -> Mapping[str, Sequence[int]]:
"""Processes one input_tensor example.
Args:
input_tensor: A tensor with shape (1, None) of a utf-8 encoded string.
Returns:
A dictionary of lists all with shape (1, self._seq_len) containing the
keys "input_word_ids", "input_type_ids", and "input_mask".
"""
tokens = self._tokenizer.tokenize(input_tensor.numpy()[0].decode("utf-8"))
tokens = tokens[0 : (self._seq_len - 2)] # account for [CLS] and [SEP]
tokens.insert(0, "[CLS]")
@ -96,7 +105,18 @@ class BertFastTokenizer(BertTokenizer):
self._sep_id = vocab.index("[SEP]")
self._pad_id = vocab.index("[PAD]")
def process(self, input_tensor: tf.Tensor) -> Mapping[str, Sequence[int]]:
def process_fn(self, input_tensor: tf.Tensor) -> Mapping[str, tf.Tensor]:
"""Tensor implementation of the process function.
This implementation can be used within a model graph directly since it
takes in tensors and outputs tensors.
Args:
input_tensor: Input string tensor
Returns:
Dictionary of tf.Tensors.
"""
input_ids = self._tokenizer.tokenize(input_tensor).flat_values
input_ids = input_ids[: (self._seq_len - 2)]
input_ids = tf.concat(
@ -112,7 +132,20 @@ class BertFastTokenizer(BertTokenizer):
input_type_ids = tf.zeros(self._seq_len, dtype=tf.int32)
input_mask = tf.cast(input_ids != self._pad_id, dtype=tf.int32)
return {
"input_word_ids": input_ids.numpy().tolist(),
"input_type_ids": input_type_ids.numpy().tolist(),
"input_mask": input_mask.numpy().tolist(),
"input_word_ids": input_ids,
"input_type_ids": input_type_ids,
"input_mask": input_mask,
}
def process(self, input_tensor: tf.Tensor) -> Mapping[str, Sequence[int]]:
"""Processes one input_tensor example.
Args:
input_tensor: A tensor with shape (1, None) of a utf-8 encoded string.
Returns:
A dictionary of lists all with shape (1, self._seq_len) containing the
keys "input_word_ids", "input_type_ids", and "input_mask".
"""
result = self.process_fn(input_tensor)
return {k: v.numpy().tolist() for k, v in result.items()}

35
mediapipe/model_maker/python/text/text_classifier/model_with_tokenizer.py Normal file
View File

@ -0,0 +1,35 @@
# Copyright 2023 The MediaPipe Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Text classifier export module library."""
import tensorflow as tf
class ModelWithTokenizer(tf.keras.Model):
"""A model with the tokenizer included in graph for exporting to TFLite."""
def __init__(self, tokenizer, model):
super().__init__()
self._tokenizer = tokenizer
self._model = model
@tf.function(
input_signature=[
tf.TensorSpec(shape=[None], dtype=tf.string, name="input")
]
)
def call(self, input_tensor):
x = self._tokenizer.process_fn(input_tensor)
x = {k: tf.expand_dims(v, axis=0) for k, v in x.items()}
x = self._model(x)
return x

105
mediapipe/model_maker/python/text/text_classifier/model_with_tokenizer_test.py Normal file
View File

@ -0,0 +1,105 @@
# Copyright 2022 The MediaPipe Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import tempfile
from unittest import mock as unittest_mock
import tensorflow as tf
import tensorflow_hub
from mediapipe.model_maker.python.core.utils import hub_loader
from mediapipe.model_maker.python.text.text_classifier import bert_tokenizer
from mediapipe.model_maker.python.text.text_classifier import model_spec
from mediapipe.model_maker.python.text.text_classifier import model_with_tokenizer
class BertTokenizerTest(tf.test.TestCase):
_SEQ_LEN = 128
def setUp(self):
super().setUp()
# Mock tempfile.gettempdir() to be unique for each test to avoid race
# condition when downloading model since these tests may run in parallel.
mock_gettempdir = unittest_mock.patch.object(
tempfile,
"gettempdir",
return_value=self.create_tempdir(),
autospec=True,
)
self.mock_gettempdir = mock_gettempdir.start()
self.addCleanup(mock_gettempdir.stop)
self._ms = model_spec.SupportedModels.MOBILEBERT_CLASSIFIER.value()
self._tokenizer = self._create_tokenizer()
self._model = self._create_model()
def _create_tokenizer(self):
vocab_file = os.path.join(
tensorflow_hub.resolve(self._ms.get_path()), "assets", "vocab.txt"
)
return bert_tokenizer.BertFastTokenizer(vocab_file, True, self._SEQ_LEN)
def _create_model(self):
encoder_inputs = dict(
input_word_ids=tf.keras.layers.Input(
shape=(self._SEQ_LEN,),
dtype=tf.int32,
name="input_word_ids",
),
input_mask=tf.keras.layers.Input(
shape=(self._SEQ_LEN,),
dtype=tf.int32,
name="input_mask",
),
input_type_ids=tf.keras.layers.Input(
shape=(self._SEQ_LEN,),
dtype=tf.int32,
name="input_type_ids",
),
)
renamed_inputs = dict(
input_ids=encoder_inputs["input_word_ids"],
input_mask=encoder_inputs["input_mask"],
segment_ids=encoder_inputs["input_type_ids"],
)
encoder = hub_loader.HubKerasLayerV1V2(
self._ms.get_path(),
signature="tokens",
output_key="pooled_output",
trainable=True,
)
pooled_output = encoder(renamed_inputs)
output = tf.keras.layers.Dropout(rate=0.1)(pooled_output)
initializer = tf.keras.initializers.TruncatedNormal(stddev=0.02)
output = tf.keras.layers.Dense(
2,
kernel_initializer=initializer,
name="output",
activation="softmax",
dtype=tf.float32,
)(output)
return tf.keras.Model(inputs=encoder_inputs, outputs=output)
def test_model_with_tokenizer(self):
model = model_with_tokenizer.ModelWithTokenizer(
self._tokenizer, self._model
)
output = model(tf.constant(["Example input".encode("utf-8")]))
self.assertAllEqual(output.shape, (1, 2))
self.assertEqual(tf.reduce_sum(output), 1)
if __name__ == "__main__":
tf.test.main()

4
mediapipe/model_maker/python/text/text_classifier/preprocessor.py
View File

@ -368,6 +368,10 @@ class BertClassifierPreprocessor:
tfrecord_cache_files=tfrecord_cache_files,
)
@property
def tokenizer(self) -> bert_tokenizer.BertTokenizer:
return self._tokenizer
TextClassifierPreprocessor = Union[
BertClassifierPreprocessor, AverageWordEmbeddingClassifierPreprocessor

55
mediapipe/model_maker/python/text/text_classifier/text_classifier.py
View File

@ -29,10 +29,12 @@ from mediapipe.model_maker.python.core.utils import loss_functions
from mediapipe.model_maker.python.core.utils import metrics
from mediapipe.model_maker.python.core.utils import model_util
from mediapipe.model_maker.python.core.utils import quantization
from mediapipe.model_maker.python.text.text_classifier import bert_tokenizer
from mediapipe.model_maker.python.text.text_classifier import dataset as text_ds
from mediapipe.model_maker.python.text.text_classifier import hyperparameters as hp
from mediapipe.model_maker.python.text.text_classifier import model_options as mo
from mediapipe.model_maker.python.text.text_classifier import model_spec as ms
from mediapipe.model_maker.python.text.text_classifier import model_with_tokenizer
from mediapipe.model_maker.python.text.text_classifier import preprocessor
from mediapipe.model_maker.python.text.text_classifier import text_classifier_options
from mediapipe.tasks.python.metadata.metadata_writers import metadata_writer
@ -620,3 +622,56 @@ class _BertClassifier(TextClassifier):
ids_name=self._model_spec.tflite_input_name["ids"],
mask_name=self._model_spec.tflite_input_name["mask"],
segment_name=self._model_spec.tflite_input_name["segment_ids"])
def export_model_with_tokenizer(
self,
model_name: str = "model_with_tokenizer.tflite",
quantization_config: Optional[quantization.QuantizationConfig] = None,
):
"""Converts and saves the model to a TFLite file with the tokenizer.
Note that unlike the export_model method, this export method will include
a FastBertTokenizer in the TFLite graph. The resulting TFLite will not have
metadata information to use with MediaPipe Tasks, but can be run directly
using TFLite Inference: https://www.tensorflow.org/lite/guide/inference
For more information on the tokenizer, see:
https://www.tensorflow.org/text/api_docs/python/text/FastBertTokenizer
Args:
model_name: File name to save TFLite model with tokenizer. The full export
path is {self._hparams.export_dir}/{model_name}.
quantization_config: The configuration for model quantization.
"""
tf.io.gfile.makedirs(self._hparams.export_dir)
tflite_file = os.path.join(self._hparams.export_dir, model_name)
if (
self._hparams.tokenizer
!= bert_tokenizer.SupportedBertTokenizers.FAST_BERT_TOKENIZER
):
print(
f"WARNING: This model was trained with {self._hparams.tokenizer} "
"tokenizer, but the exported model with tokenizer will have a "
f"{bert_tokenizer.SupportedBertTokenizers.FAST_BERT_TOKENIZER} "
"tokenizer."
)
tokenizer = bert_tokenizer.BertFastTokenizer(
vocab_file=self._text_preprocessor.get_vocab_file(),
do_lower_case=self._model_spec.do_lower_case,
seq_len=self._model_options.seq_len,
)
else:
tokenizer = self._text_preprocessor.tokenizer
model = model_with_tokenizer.ModelWithTokenizer(tokenizer, self._model)
model(tf.constant(["Example input data".encode("utf-8")])) # build model
saved_model_file = os.path.join(
self._hparams.export_dir, "saved_model_with_tokenizer"
)
model.save(saved_model_file)
tflite_model = model_util.convert_to_tflite_from_file(
saved_model_file,
quantization_config=quantization_config,
allow_custom_ops=True,
)
model_util.save_tflite(tflite_model, tflite_file)

6
mediapipe/model_maker/python/text/text_classifier/text_classifier_test.py
View File

@ -149,6 +149,12 @@ class TextClassifierTest(tf.test.TestCase, parameterized.TestCase):
output_metadata_file, self._BERT_CLASSIFIER_JSON_FILE, shallow=False
)
)
bert_classifier.export_model_with_tokenizer()
output_tflite_with_tokenizer_file = os.path.join(
options.hparams.export_dir, 'model_with_tokenizer.tflite'
)
self.assertTrue(os.path.exists(output_tflite_with_tokenizer_file))
self.assertGreater(os.path.getsize(output_tflite_with_tokenizer_file), 0)
def test_label_mismatch(self):
options = text_classifier.TextClassifierOptions(

6
mediapipe/tasks/ios/common/utils/sources/MPPCommonUtils.mm
View File

@ -59,12 +59,12 @@ using absl::StatusCode;
return NULL;
}
void *allocedMemory = malloc(memSize);
if (!allocedMemory) {
void *allocatedMemory = malloc(memSize);
if (!allocatedMemory) {
exit(-1);
}
return allocedMemory;
return allocatedMemory;
}
+ (BOOL)checkCppError:(const absl::Status &)status toError:(NSError *_Nullable *)error {

2
mediapipe/tasks/ios/test/vision/core/MPPImageTests.m
View File

@ -82,7 +82,7 @@ static NSString *const kExpectedErrorDomain = @"com.google.mediapipe.tasks";
AssertEqualErrors(error, expectedError);
}
- (void)testInitWithImageSuceeds {
- (void)testInitWithImageSucceeds {
MPPImage *mppImage = [[MPPImage alloc] initWithUIImage:self.image error:nil];
[self assertMPPImage:mppImage
hasSourceType:MPPImageSourceTypeImage

28
mediapipe/tasks/ios/test/vision/face_detector/MPPFaceDetectorTests.mm
View File

@ -109,7 +109,7 @@ static const float kKeypointErrorThreshold = 1e-2;
NSError *error;
MPPImage *mppImage = [self imageWithFileInfo:kCatImage];
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectInImage:mppImage error:&error];
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectImage:mppImage error:&error];
XCTAssertNil(error);
XCTAssertNotNil(faceDetectorResult);
XCTAssertEqual(faceDetectorResult.detections.count, 0);
@ -125,7 +125,7 @@ static const float kKeypointErrorThreshold = 1e-2;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
for (int i = 0; i < 3; i++) {
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectInVideoFrame:image
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectVideoFrame:image
timestampInMilliseconds:i
error:nil];
[self assertFaceDetectorResult:faceDetectorResult
@ -141,7 +141,7 @@ static const float kKeypointErrorThreshold = 1e-2;
MPPImage *image = [self imageWithFileInfo:kPortraitRotatedImage];
for (int i = 0; i < 3; i++) {
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectInVideoFrame:image
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectVideoFrame:image
timestampInMilliseconds:i
error:nil];
[self assertFaceDetectorResult:faceDetectorResult
@ -181,7 +181,7 @@ static const float kKeypointErrorThreshold = 1e-2;
};
for (int i = 0; i < iterationCount; i++) {
XCTAssertTrue([faceDetector detectAsyncInImage:image timestampInMilliseconds:i error:nil]);
XCTAssertTrue([faceDetector detectAsyncImage:image timestampInMilliseconds:i error:nil]);
}
NSTimeInterval timeout = 0.5f;
@ -205,10 +205,10 @@ static const float kKeypointErrorThreshold = 1e-2;
};
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
XCTAssertTrue([faceDetector detectAsyncInImage:image timestampInMilliseconds:1 error:nil]);
XCTAssertTrue([faceDetector detectAsyncImage:image timestampInMilliseconds:1 error:nil]);
NSError *error;
XCTAssertFalse([faceDetector detectAsyncInImage:image timestampInMilliseconds:0 error:&error]);
XCTAssertFalse([faceDetector detectAsyncImage:image timestampInMilliseconds:0 error:&error]);
NSError *expectedError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -274,7 +274,7 @@ static const float kKeypointErrorThreshold = 1e-2;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
NSError *liveStreamApiCallError;
XCTAssertFalse([faceDetector detectAsyncInImage:image
XCTAssertFalse([faceDetector detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamApiCallError]);
@ -288,7 +288,7 @@ static const float kKeypointErrorThreshold = 1e-2;
AssertEqualErrors(liveStreamApiCallError, expectedLiveStreamApiCallError);
NSError *videoApiCallError;
XCTAssertFalse([faceDetector detectInVideoFrame:image
XCTAssertFalse([faceDetector detectVideoFrame:image
timestampInMilliseconds:0
error:&videoApiCallError]);
@ -312,7 +312,7 @@ static const float kKeypointErrorThreshold = 1e-2;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
NSError *liveStreamApiCallError;
XCTAssertFalse([faceDetector detectAsyncInImage:image
XCTAssertFalse([faceDetector detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamApiCallError]);
@ -326,7 +326,7 @@ static const float kKeypointErrorThreshold = 1e-2;
AssertEqualErrors(liveStreamApiCallError, expectedLiveStreamApiCallError);
NSError *imageApiCallError;
XCTAssertFalse([faceDetector detectInImage:image error:&imageApiCallError]);
XCTAssertFalse([faceDetector detectImage:image error:&imageApiCallError]);
NSError *expectedImageApiCallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -350,7 +350,7 @@ static const float kKeypointErrorThreshold = 1e-2;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
NSError *imageApiCallError;
XCTAssertFalse([faceDetector detectInImage:image error:&imageApiCallError]);
XCTAssertFalse([faceDetector detectImage:image error:&imageApiCallError]);
NSError *expectedImageApiCallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -362,7 +362,7 @@ static const float kKeypointErrorThreshold = 1e-2;
AssertEqualErrors(imageApiCallError, expectedImageApiCallError);
NSError *videoApiCallError;
XCTAssertFalse([faceDetector detectInVideoFrame:image
XCTAssertFalse([faceDetector detectVideoFrame:image
timestampInMilliseconds:0
error:&videoApiCallError]);
@ -407,7 +407,7 @@ static const float kKeypointErrorThreshold = 1e-2;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
for (int i = 0; i < iterationCount; i++) {
XCTAssertTrue([faceDetector detectAsyncInImage:image timestampInMilliseconds:i error:nil]);
XCTAssertTrue([faceDetector detectAsyncImage:image timestampInMilliseconds:i error:nil]);
}
NSTimeInterval timeout = 0.5f;
@ -503,7 +503,7 @@ static const float kKeypointErrorThreshold = 1e-2;
usingFaceDetector:(MPPFaceDetector *)faceDetector
containsExpectedKeypoints:(NSArray<NSArray *> *)expectedKeypoints {
NSError *error;
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectInImage:mppImage error:&error];
MPPFaceDetectorResult *faceDetectorResult = [faceDetector detectImage:mppImage error:&error];
XCTAssertNil(error);
XCTAssertNotNil(faceDetectorResult);
[self assertFaceDetectorResult:faceDetectorResult containsExpectedKeypoints:expectedKeypoints];

26
mediapipe/tasks/ios/test/vision/face_landmarker/MPPFaceLandmarkerTests.mm
View File

@ -137,7 +137,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
NSError *error;
MPPImage *mppImage = [self imageWithFileInfo:kCatImage];
MPPFaceLandmarkerResult *faceLandmarkerResult = [faceLandmarker detectInImage:mppImage
MPPFaceLandmarkerResult *faceLandmarkerResult = [faceLandmarker detectImage:mppImage
error:&error];
XCTAssertNil(error);
XCTAssertNotNil(faceLandmarkerResult);
@ -158,7 +158,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
NSArray<MPPNormalizedLandmark *> *expectedLandmarks =
[MPPFaceLandmarkerTests expectedLandmarksFromFileInfo:kPortraitExpectedLandmarksName];
for (int i = 0; i < 3; i++) {
MPPFaceLandmarkerResult *faceLandmarkerResult = [faceLandmarker detectInVideoFrame:image
MPPFaceLandmarkerResult *faceLandmarkerResult = [faceLandmarker detectVideoFrame:image
timestampInMilliseconds:i
error:nil];
[self assertFaceLandmarkerResult:faceLandmarkerResult
@ -200,7 +200,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
};
for (int i = 0; i < iterationCount; i++) {
XCTAssertTrue([faceLandmarker detectAsyncInImage:image timestampInMilliseconds:i error:nil]);
XCTAssertTrue([faceLandmarker detectAsyncImage:image timestampInMilliseconds:i error:nil]);
}
NSTimeInterval timeout = 0.5f;
@ -224,10 +224,10 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
};
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
XCTAssertTrue([faceLandmarker detectAsyncInImage:image timestampInMilliseconds:1 error:nil]);
XCTAssertTrue([faceLandmarker detectAsyncImage:image timestampInMilliseconds:1 error:nil]);
NSError *error;
XCTAssertFalse([faceLandmarker detectAsyncInImage:image timestampInMilliseconds:0 error:&error]);
XCTAssertFalse([faceLandmarker detectAsyncImage:image timestampInMilliseconds:0 error:&error]);
NSError *expectedError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -292,7 +292,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
NSError *liveStreamAPICallError;
XCTAssertFalse([faceLandmarker detectAsyncInImage:image
XCTAssertFalse([faceLandmarker detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamAPICallError]);
@ -306,7 +306,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
AssertEqualErrors(liveStreamAPICallError, expectedLiveStreamAPICallError);
NSError *videoAPICallError;
XCTAssertFalse([faceLandmarker detectInVideoFrame:image
XCTAssertFalse([faceLandmarker detectVideoFrame:image
timestampInMilliseconds:0
error:&videoAPICallError]);
@ -329,7 +329,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
NSError *liveStreamAPICallError;
XCTAssertFalse([faceLandmarker detectAsyncInImage:image
XCTAssertFalse([faceLandmarker detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamAPICallError]);
@ -343,7 +343,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
AssertEqualErrors(liveStreamAPICallError, expectedLiveStreamAPICallError);
NSError *imageAPICallError;
XCTAssertFalse([faceLandmarker detectInImage:image error:&imageAPICallError]);
XCTAssertFalse([faceLandmarker detectImage:image error:&imageAPICallError]);
NSError *expectedImageAPICallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -365,7 +365,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
MPPImage *image = [self imageWithFileInfo:kPortraitImage];
NSError *imageAPICallError;
XCTAssertFalse([faceLandmarker detectInImage:image error:&imageAPICallError]);
XCTAssertFalse([faceLandmarker detectImage:image error:&imageAPICallError]);
NSError *expectedImageAPICallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -377,7 +377,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
AssertEqualErrors(imageAPICallError, expectedImageAPICallError);
NSError *videoAPICallError;
XCTAssertFalse([faceLandmarker detectInVideoFrame:image
XCTAssertFalse([faceLandmarker detectVideoFrame:image
timestampInMilliseconds:0
error:&videoAPICallError]);
@ -484,7 +484,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
}
}
if (expectedTransformationMatrix == NULL) {
if (expectedTransformationMatrix == nullptr) {
XCTAssertEqualObjects(faceLandmarkerResult.facialTransformationMatrixes, [NSArray array]);
} else {
MPPTransformMatrix *actualTransformationMatrix =
@ -539,7 +539,7 @@ constexpr float kFacialTransformationMatrixErrorThreshold = 0.2f;
MPPImage *mppImage = [self imageWithFileInfo:fileInfo];
NSError *error;
MPPFaceLandmarkerResult *faceLandmarkerResult = [faceLandmarker detectInImage:mppImage
MPPFaceLandmarkerResult *faceLandmarkerResult = [faceLandmarker detectImage:mppImage
error:&error];
XCTAssertNil(error);
XCTAssertNotNil(faceLandmarkerResult);

4
mediapipe/tasks/ios/test/vision/face_landmarker/utils/BUILD
View File

@ -25,7 +25,7 @@ TFL_DISABLED_SANITIZER_TAGS = [
]
objc_library(
name = "MPPFaceLandmarkeResultHelpersTestLibary",
name = "MPPFaceLandmarkeResultHelpersTestLibrary",
testonly = 1,
srcs = ["sources/MPPFaceLandmarkerResult+HelpersTests.mm"],
copts = [
@ -50,6 +50,6 @@ ios_unit_test(
runner = tflite_ios_lab_runner("IOS_LATEST"),
tags = TFL_DEFAULT_TAGS + TFL_DISABLED_SANITIZER_TAGS,
deps = [
":MPPFaceLandmarkeResultHelpersTestLibary",
":MPPFaceLandmarkeResultHelpersTestLibrary",
],
)

32
mediapipe/tasks/ios/test/vision/hand_landmarker/MPPHandLandmarkerTests.m
View File

@ -208,10 +208,10 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
return image;
}
- (MPPHandLandmarkerResult *)detectInImageWithFileInfo:(ResourceFileInfo *)imageFileInfo
- (MPPHandLandmarkerResult *)detectImageWithFileInfo:(ResourceFileInfo *)imageFileInfo
usingHandLandmarker:(MPPHandLandmarker *)handLandmarker {
MPPImage *mppImage = [self imageWithFileInfo:imageFileInfo];
MPPHandLandmarkerResult *handLandmarkerResult = [handLandmarker detectInImage:mppImage error:nil];
MPPHandLandmarkerResult *handLandmarkerResult = [handLandmarker detectImage:mppImage error:nil];
XCTAssertNotNil(handLandmarkerResult);
return handLandmarkerResult;
@ -221,7 +221,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
usingHandLandmarker:(MPPHandLandmarker *)handLandmarker
approximatelyEqualsHandLandmarkerResult:
(MPPHandLandmarkerResult *)expectedHandLandmarkerResult {
MPPHandLandmarkerResult *handLandmarkerResult = [self detectInImageWithFileInfo:fileInfo
MPPHandLandmarkerResult *handLandmarkerResult = [self detectImageWithFileInfo:fileInfo
usingHandLandmarker:handLandmarker];
[self assertHandLandmarkerResult:handLandmarkerResult
isApproximatelyEqualToExpectedResult:expectedHandLandmarkerResult];
@ -249,7 +249,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPHandLandmarker *handLandmarker =
[self createHandLandmarkerWithOptionsSucceeds:handLandmarkerOptions];
MPPHandLandmarkerResult *handLandmarkerResult = [self detectInImageWithFileInfo:kNoHandsImage
MPPHandLandmarkerResult *handLandmarkerResult = [self detectImageWithFileInfo:kNoHandsImage
usingHandLandmarker:handLandmarker];
AssertHandLandmarkerResultIsEmpty(handLandmarkerResult);
}
@ -264,7 +264,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPHandLandmarker *handLandmarker =
[self createHandLandmarkerWithOptionsSucceeds:handLandmarkerOptions];
MPPHandLandmarkerResult *handLandmarkerResult = [self detectInImageWithFileInfo:kTwoHandsImage
MPPHandLandmarkerResult *handLandmarkerResult = [self detectImageWithFileInfo:kTwoHandsImage
usingHandLandmarker:handLandmarker];
XCTAssertTrue(handLandmarkerResult.handedness.count == numHands);
@ -280,7 +280,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *mppImage = [self imageWithFileInfo:kPointingUpRotatedImage
orientation:UIImageOrientationRight];
MPPHandLandmarkerResult *handLandmarkerResult = [handLandmarker detectInImage:mppImage error:nil];
MPPHandLandmarkerResult *handLandmarkerResult = [handLandmarker detectImage:mppImage error:nil];
[self assertHandLandmarkerResult:handLandmarkerResult
isApproximatelyEqualToExpectedResult:[MPPHandLandmarkerTests
@ -339,7 +339,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kThumbUpImage];
NSError *liveStreamApiCallError;
XCTAssertFalse([handLandmarker detectAsyncInImage:image
XCTAssertFalse([handLandmarker detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamApiCallError]);
@ -354,7 +354,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
AssertEqualErrors(liveStreamApiCallError, expectedLiveStreamApiCallError);
NSError *videoApiCallError;
XCTAssertFalse([handLandmarker detectInVideoFrame:image
XCTAssertFalse([handLandmarker detectVideoFrame:image
timestampInMilliseconds:0
error:&videoApiCallError]);
@ -378,7 +378,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kThumbUpImage];
NSError *liveStreamApiCallError;
XCTAssertFalse([handLandmarker detectAsyncInImage:image
XCTAssertFalse([handLandmarker detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamApiCallError]);
@ -393,7 +393,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
AssertEqualErrors(liveStreamApiCallError, expectedLiveStreamApiCallError);
NSError *imageApiCallError;
XCTAssertFalse([handLandmarker detectInImage:image error:&imageApiCallError]);
XCTAssertFalse([handLandmarker detectImage:image error:&imageApiCallError]);
NSError *expectedImageApiCallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -416,7 +416,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kThumbUpImage];
NSError *imageApiCallError;
XCTAssertFalse([handLandmarker detectInImage:image error:&imageApiCallError]);
XCTAssertFalse([handLandmarker detectImage:image error:&imageApiCallError]);
NSError *expectedImageApiCallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -428,7 +428,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
AssertEqualErrors(imageApiCallError, expectedImageApiCallError);
NSError *videoApiCallError;
XCTAssertFalse([handLandmarker detectInVideoFrame:image
XCTAssertFalse([handLandmarker detectVideoFrame:image
timestampInMilliseconds:0
error:&videoApiCallError]);
@ -452,7 +452,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kThumbUpImage];
for (int i = 0; i < 3; i++) {
MPPHandLandmarkerResult *handLandmarkerResult = [handLandmarker detectInVideoFrame:image
MPPHandLandmarkerResult *handLandmarkerResult = [handLandmarker detectVideoFrame:image
timestampInMilliseconds:i
error:nil];
[self assertHandLandmarkerResult:handLandmarkerResult
@ -480,10 +480,10 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kThumbUpImage];
XCTAssertTrue([handLandmarker detectAsyncInImage:image timestampInMilliseconds:1 error:nil]);
XCTAssertTrue([handLandmarker detectAsyncImage:image timestampInMilliseconds:1 error:nil]);
NSError *error;
XCTAssertFalse([handLandmarker detectAsyncInImage:image timestampInMilliseconds:0 error:&error]);
XCTAssertFalse([handLandmarker detectAsyncImage:image timestampInMilliseconds:0 error:&error]);
NSError *expectedError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -533,7 +533,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kThumbUpImage];
for (int i = 0; i < iterationCount; i++) {
XCTAssertTrue([handLandmarker detectAsyncInImage:image timestampInMilliseconds:i error:nil]);
XCTAssertTrue([handLandmarker detectAsyncImage:image timestampInMilliseconds:i error:nil]);
}
NSTimeInterval timeout = 0.5f;

22
mediapipe/tasks/ios/test/vision/object_detector/MPPObjectDetectorTests.m
View File

@ -194,7 +194,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
usingObjectDetector:(MPPObjectDetector *)objectDetector
maxResults:(NSInteger)maxResults
equalsObjectDetectorResult:(MPPObjectDetectorResult *)expectedObjectDetectorResult {
MPPObjectDetectorResult *ObjectDetectorResult = [objectDetector detectInImage:mppImage error:nil];
MPPObjectDetectorResult *ObjectDetectorResult = [objectDetector detectImage:mppImage error:nil];
[self assertObjectDetectorResult:ObjectDetectorResult
isEqualToExpectedResult:expectedObjectDetectorResult
@ -495,7 +495,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kCatsAndDogsImage];
NSError *liveStreamApiCallError;
XCTAssertFalse([objectDetector detectAsyncInImage:image
XCTAssertFalse([objectDetector detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamApiCallError]);
@ -510,7 +510,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
AssertEqualErrors(liveStreamApiCallError, expectedLiveStreamApiCallError);
NSError *videoApiCallError;
XCTAssertFalse([objectDetector detectInVideoFrame:image
XCTAssertFalse([objectDetector detectVideoFrame:image
timestampInMilliseconds:0
error:&videoApiCallError]);
@ -533,7 +533,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kCatsAndDogsImage];
NSError *liveStreamApiCallError;
XCTAssertFalse([objectDetector detectAsyncInImage:image
XCTAssertFalse([objectDetector detectAsyncImage:image
timestampInMilliseconds:0
error:&liveStreamApiCallError]);
@ -548,7 +548,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
AssertEqualErrors(liveStreamApiCallError, expectedLiveStreamApiCallError);
NSError *imageApiCallError;
XCTAssertFalse([objectDetector detectInImage:image error:&imageApiCallError]);
XCTAssertFalse([objectDetector detectImage:image error:&imageApiCallError]);
NSError *expectedImageApiCallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -571,7 +571,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kCatsAndDogsImage];
NSError *imageApiCallError;
XCTAssertFalse([objectDetector detectInImage:image error:&imageApiCallError]);
XCTAssertFalse([objectDetector detectImage:image error:&imageApiCallError]);
NSError *expectedImageApiCallError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -583,7 +583,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
AssertEqualErrors(imageApiCallError, expectedImageApiCallError);
NSError *videoApiCallError;
XCTAssertFalse([objectDetector detectInVideoFrame:image
XCTAssertFalse([objectDetector detectVideoFrame:image
timestampInMilliseconds:0
error:&videoApiCallError]);
@ -610,7 +610,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kCatsAndDogsImage];
for (int i = 0; i < 3; i++) {
MPPObjectDetectorResult *ObjectDetectorResult = [objectDetector detectInVideoFrame:image
MPPObjectDetectorResult *ObjectDetectorResult = [objectDetector detectVideoFrame:image
timestampInMilliseconds:i
error:nil];
@ -643,10 +643,10 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kCatsAndDogsImage];
XCTAssertTrue([objectDetector detectAsyncInImage:image timestampInMilliseconds:1 error:nil]);
XCTAssertTrue([objectDetector detectAsyncImage:image timestampInMilliseconds:1 error:nil]);
NSError *error;
XCTAssertFalse([objectDetector detectAsyncInImage:image timestampInMilliseconds:0 error:&error]);
XCTAssertFalse([objectDetector detectAsyncImage:image timestampInMilliseconds:0 error:&error]);
NSError *expectedError =
[NSError errorWithDomain:kExpectedErrorDomain
@ -702,7 +702,7 @@ static NSString *const kLiveStreamTestsDictExpectationKey = @"expectation";
MPPImage *image = [self imageWithFileInfo:kCatsAndDogsImage];
for (int i = 0; i < iterationCount; i++) {
XCTAssertTrue([objectDetector detectAsyncInImage:image timestampInMilliseconds:i error:nil]);
XCTAssertTrue([objectDetector detectAsyncImage:image timestampInMilliseconds:i error:nil]);
}
NSTimeInterval timeout = 0.5f;

6
mediapipe/tasks/ios/vision/face_detector/sources/MPPFaceDetector.h
View File

@ -100,7 +100,7 @@ NS_SWIFT_NAME(FaceDetector)
* system, i.e. in `[0,image_width) x [0,image_height)`, which are the dimensions of the underlying
* image data.
*/
- (nullable MPPFaceDetectorResult *)detectInImage:(MPPImage *)image
- (nullable MPPFaceDetectorResult *)detectImage:(MPPImage *)image
error:(NSError **)error NS_SWIFT_NAME(detect(image:));
/**
@ -127,7 +127,7 @@ NS_SWIFT_NAME(FaceDetector)
* system, i.e. in `[0,image_width) x [0,image_height)`, which are the dimensions of the underlying
* image data.
*/
- (nullable MPPFaceDetectorResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPFaceDetectorResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detect(videoFrame:timestampInMilliseconds:));
@ -165,7 +165,7 @@ NS_SWIFT_NAME(FaceDetector)
*
* @return `true` if the image was sent to the task successfully, otherwise `false`.
*/
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detectAsync(image:timestampInMilliseconds:));

6
mediapipe/tasks/ios/vision/face_detector/sources/MPPFaceDetector.mm
View File

@ -130,13 +130,13 @@ static NSString *const kTaskName = @"faceDetector";
return [self initWithOptions:options error:error];
}
- (nullable MPPFaceDetectorResult *)detectInImage:(MPPImage *)image error:(NSError **)error {
- (nullable MPPFaceDetectorResult *)detectImage:(MPPImage *)image error:(NSError **)error {
std::optional<PacketMap> outputPacketMap = [_visionTaskRunner processImage:image error:error];
return [MPPFaceDetector faceDetectorResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (nullable MPPFaceDetectorResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPFaceDetectorResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
std::optional<PacketMap> outputPacketMap =
@ -147,7 +147,7 @@ static NSString *const kTaskName = @"faceDetector";
return [MPPFaceDetector faceDetectorResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
return [_visionTaskRunner processLiveStreamImage:image

6
mediapipe/tasks/ios/vision/face_landmarker/sources/MPPFaceLandmarker.h
View File

@ -71,7 +71,7 @@ NS_SWIFT_NAME(FaceLandmarker)
* @return An `FaceLandmarkerResult` that contains a list of landmarks. `nil` if there is an error
* in initializing the face landmaker.
*/
- (nullable MPPFaceLandmarkerResult *)detectInImage:(MPPImage *)image
- (nullable MPPFaceLandmarkerResult *)detectImage:(MPPImage *)image
error:(NSError **)error NS_SWIFT_NAME(detect(image:));
/**
@ -95,7 +95,7 @@ NS_SWIFT_NAME(FaceLandmarker)
* @return An `FaceLandmarkerResult` that contains a list of landmarks. `nil` if there is an
* error in initializing the face landmaker.
*/
- (nullable MPPFaceLandmarkerResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPFaceLandmarkerResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detect(videoFrame:timestampInMilliseconds:));
@ -132,7 +132,7 @@ NS_SWIFT_NAME(FaceLandmarker)
*
* @return `true` if the image was sent to the task successfully, otherwise `false`.
*/
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detectAsync(image:timestampInMilliseconds:));

6
mediapipe/tasks/ios/vision/face_landmarker/sources/MPPFaceLandmarker.mm
View File

@ -154,13 +154,13 @@ static NSString *const kTaskName = @"faceLandmarker";
return [self initWithOptions:options error:error];
}
- (nullable MPPFaceLandmarkerResult *)detectInImage:(MPPImage *)image error:(NSError **)error {
- (nullable MPPFaceLandmarkerResult *)detectImage:(MPPImage *)image error:(NSError **)error {
std::optional<PacketMap> outputPacketMap = [_visionTaskRunner processImage:image error:error];
return [MPPFaceLandmarker faceLandmarkerResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (nullable MPPFaceLandmarkerResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPFaceLandmarkerResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
std::optional<PacketMap> outputPacketMap =
@ -171,7 +171,7 @@ static NSString *const kTaskName = @"faceLandmarker";
return [MPPFaceLandmarker faceLandmarkerResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
return [_visionTaskRunner processLiveStreamImage:image

4
mediapipe/tasks/ios/vision/face_stylizer/sources/MPPFaceStylizer.h
View File

@ -68,7 +68,7 @@ NS_SWIFT_NAME(FaceStylizer)
* @return A `FaceStylizerResult` that contains the stylized image of the most visible face. The
* returned image is copied. The stylized output image size is the same as the model output
* size. The `stylizedImage` of the `FaceStylizerResult` is `nil` if there is no face detected in
* the imput image. `FaceStylizerResult` is `nil` if there is an error in initializing the face
* the input image. `FaceStylizerResult` is `nil` if there is an error in initializing the face
* stylizer.
*/
- (nullable MPPFaceStylizerResult *)stylizeImage:(MPPImage *)image
@ -92,7 +92,7 @@ NS_SWIFT_NAME(FaceStylizer)
*
* @param image The `MPImage` on which face stylization is to be performed.
* @param completionHandler A block to be invoked with the results of performing face stylization on
* the imput image. The block takes two arguments, the optional `FaceStylizerResult` that contains
* the input image. The block takes two arguments, the optional `FaceStylizerResult` that contains
* the zero-copied stylized image if face stylization was successful and an optional error populated
* upon failure. The lifetime of the stylized image is only guaranteed for the duration of the
* block.

6
mediapipe/tasks/ios/vision/hand_landmarker/sources/MPPHandLandmarker.h
View File

@ -146,7 +146,7 @@ NS_SWIFT_NAME(HandLandmarker)
* @return An `HandLandmarkerResult` object that contains the hand hand landmarks detection
* results.
*/
- (nullable MPPHandLandmarkerResult *)detectInImage:(MPPImage *)image
- (nullable MPPHandLandmarkerResult *)detectImage:(MPPImage *)image
error:(NSError **)error NS_SWIFT_NAME(detect(image:));
/**
@ -176,7 +176,7 @@ NS_SWIFT_NAME(HandLandmarker)
* @return An `HandLandmarkerResult` object that contains the hand hand landmarks detection
* results.
*/
- (nullable MPPHandLandmarkerResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPHandLandmarkerResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detect(videoFrame:timestampInMilliseconds:));
@ -216,7 +216,7 @@ NS_SWIFT_NAME(HandLandmarker)
*
* @return `YES` if the image was sent to the task successfully, otherwise `NO`.
*/
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detectAsync(image:timestampInMilliseconds:));

6
mediapipe/tasks/ios/vision/hand_landmarker/sources/MPPHandLandmarker.mm
View File

@ -140,13 +140,13 @@ static NSString *const kTaskName = @"handLandmarker";
return [self initWithOptions:options error:error];
}
- (nullable MPPHandLandmarkerResult *)detectInImage:(MPPImage *)image error:(NSError **)error {
- (nullable MPPHandLandmarkerResult *)detectImage:(MPPImage *)image error:(NSError **)error {
std::optional<PacketMap> outputPacketMap = [_visionTaskRunner processImage:image error:error];
return [MPPHandLandmarker handLandmarkerResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (nullable MPPHandLandmarkerResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPHandLandmarkerResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
std::optional<PacketMap> outputPacketMap =
@ -157,7 +157,7 @@ static NSString *const kTaskName = @"handLandmarker";
return [MPPHandLandmarker handLandmarkerResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
return [_visionTaskRunner processLiveStreamImage:image

6
mediapipe/tasks/ios/vision/object_detector/sources/MPPObjectDetector.h
View File

@ -112,7 +112,7 @@ NS_SWIFT_NAME(ObjectDetector)
* system, i.e. in `[0,image_width) x [0,image_height)`, which are the dimensions of the underlying
* image data.
*/
- (nullable MPPObjectDetectorResult *)detectInImage:(MPPImage *)image
- (nullable MPPObjectDetectorResult *)detectImage:(MPPImage *)image
error:(NSError **)error NS_SWIFT_NAME(detect(image:));
/**
@ -138,7 +138,7 @@ NS_SWIFT_NAME(ObjectDetector)
* system, i.e. in `[0,image_width) x [0,image_height)`, which are the dimensions of the underlying
* image data.
*/
- (nullable MPPObjectDetectorResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPObjectDetectorResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detect(videoFrame:timestampInMilliseconds:));
@ -176,7 +176,7 @@ NS_SWIFT_NAME(ObjectDetector)
*
* @return `true` if the image was sent to the task successfully, otherwise `false`.
*/
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error
NS_SWIFT_NAME(detectAsync(image:timestampInMilliseconds:));

6
mediapipe/tasks/ios/vision/object_detector/sources/MPPObjectDetector.mm
View File

@ -128,13 +128,13 @@ static NSString *const kTaskName = @"objectDetector";
return [self initWithOptions:options error:error];
}
- (nullable MPPObjectDetectorResult *)detectInImage:(MPPImage *)image error:(NSError **)error {
- (nullable MPPObjectDetectorResult *)detectImage:(MPPImage *)image error:(NSError **)error {
std::optional<PacketMap> outputPacketMap = [_visionTaskRunner processImage:image error:error];
return [MPPObjectDetector objectDetectorResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (nullable MPPObjectDetectorResult *)detectInVideoFrame:(MPPImage *)image
- (nullable MPPObjectDetectorResult *)detectVideoFrame:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
std::optional<PacketMap> outputPacketMap =
@ -145,7 +145,7 @@ static NSString *const kTaskName = @"objectDetector";
return [MPPObjectDetector objectDetectorResultWithOptionalOutputPacketMap:outputPacketMap];
}
- (BOOL)detectAsyncInImage:(MPPImage *)image
- (BOOL)detectAsyncImage:(MPPImage *)image
timestampInMilliseconds:(NSInteger)timestampInMilliseconds
error:(NSError **)error {
return [_visionTaskRunner processLiveStreamImage:image

2
mediapipe/tasks/java/com/google/mediapipe/tasks/core/TaskOptions.java
View File

@ -25,7 +25,7 @@ import com.google.protobuf.ByteString;
/**
* MediaPipe Tasks options base class. Any MediaPipe task-specific options class should extend
* {@link TaskOptions} and implement exactly one of converTo*Proto() methods.
* {@link TaskOptions} and implement exactly one of convertTo*Proto() methods.
*/
public abstract class TaskOptions {
/**

2
mediapipe/tasks/java/com/google/mediapipe/tasks/vision/imagesegmenter/ImageSegmenter.java
View File

@ -739,7 +739,7 @@ public final class ImageSegmenter extends BaseVisionTaskApi {
@AutoValue
public abstract static class SegmentationOptions {
/** Builder fo {@link SegmentationOptions} */
/** Builder for {@link SegmentationOptions} */
@AutoValue.Builder
public abstract static class Builder {

2
mediapipe/tasks/javatests/com/google/mediapipe/tasks/vision/gesturerecognizer/GestureRecognizerTest.java
View File

@ -289,7 +289,7 @@ public class GestureRecognizerTest {
}
@Test
public void recognize_successWithPreferAlowListThanDenyList() throws Exception {
public void recognize_successWithPreferAllowListThanDenyList() throws Exception {
GestureRecognizerOptions options =
GestureRecognizerOptions.builder()
.setBaseOptions(

4
mediapipe/tasks/python/metadata/metadata_writers/metadata_writer.py
View File

@ -736,7 +736,7 @@ class MetadataWriter(object):
content is used to interpret the metadata content.
Returns:
A tuple of (model_with_metadata_in_bytes, metdata_json_content)
A tuple of (model_with_metadata_in_bytes, metadata_json_content)
"""
# Populates metadata and associated files into TFLite model buffer.
populator = metadata.MetadataPopulator.with_model_buffer(self._model_buffer)
@ -840,6 +840,6 @@ class MetadataWriterBase:
content is used to interpret the metadata content.
Returns:
A tuple of (model_with_metadata_in_bytes, metdata_json_content)
A tuple of (model_with_metadata_in_bytes, metadata_json_content)
"""
return self.writer.populate()

2
mediapipe/tasks/web/audio/audio_classifier/audio_classifier.ts
View File

@ -140,7 +140,7 @@ export class AudioClassifier extends AudioTaskRunner<AudioClassifierResult[]> {
* @param sampleRate The sample rate in Hz of the provided audio data. If not
* set, defaults to the sample rate set via `setDefaultSampleRate()` or
* `48000` if no custom default was set.
* @return The classification result of the audio datas
* @return The classification result of the audio data
*/
classify(audioData: Float32Array, sampleRate?: number):
AudioClassifierResult[] {

2
mediapipe/tasks/web/package.json
View File

@ -9,7 +9,7 @@
"import": "./__NAME___bundle.mjs",
"require": "./__NAME___bundle.cjs",
"default": "./__NAME___bundle.mjs",
"types": "./__NAME___.d.ts"
"types": "./__TYPES__"
},
"author": "mediapipe@google.com",
"license": "Apache-2.0",

4
mediapipe/tasks/web/vision/gesture_recognizer/gesture_recognizer.ts
View File

@ -342,7 +342,7 @@ export class GestureRecognizer extends VisionTaskRunner {
* Converts raw data into a landmark, and adds it to our worldLandmarks
* list.
*/
private adddJsWorldLandmarks(data: Uint8Array[]): void {
private addJsWorldLandmarks(data: Uint8Array[]): void {
for (const binaryProto of data) {
const handWorldLandmarksProto =
LandmarkList.deserializeBinary(binaryProto);
@ -396,7 +396,7 @@ export class GestureRecognizer extends VisionTaskRunner {
this.graphRunner.attachProtoVectorListener(
WORLD_LANDMARKS_STREAM, (binaryProto, timestamp) => {
this.adddJsWorldLandmarks(binaryProto);
this.addJsWorldLandmarks(binaryProto);
this.setLatestOutputTimestamp(timestamp);
});
this.graphRunner.attachEmptyPacketListener(

4
mediapipe/tasks/web/vision/hand_landmarker/hand_landmarker.ts
View File

@ -277,7 +277,7 @@ export class HandLandmarker extends VisionTaskRunner {
* Converts raw data into a world landmark, and adds it to our worldLandmarks
* list.
*/
private adddJsWorldLandmarks(data: Uint8Array[]): void {
private addJsWorldLandmarks(data: Uint8Array[]): void {
for (const binaryProto of data) {
const handWorldLandmarksProto =
LandmarkList.deserializeBinary(binaryProto);
@ -322,7 +322,7 @@ export class HandLandmarker extends VisionTaskRunner {
this.graphRunner.attachProtoVectorListener(
WORLD_LANDMARKS_STREAM, (binaryProto, timestamp) => {
this.adddJsWorldLandmarks(binaryProto);
this.addJsWorldLandmarks(binaryProto);
this.setLatestOutputTimestamp(timestamp);
});
this.graphRunner.attachEmptyPacketListener(

4
mediapipe/tasks/web/vision/pose_landmarker/pose_landmarker.ts
View File

@ -403,7 +403,7 @@ export class PoseLandmarker extends VisionTaskRunner {
* Converts raw data into a world landmark, and adds it to our
* worldLandmarks list.
*/
private adddJsWorldLandmarks(data: Uint8Array[]): void {
private addJsWorldLandmarks(data: Uint8Array[]): void {
this.worldLandmarks = [];
for (const binaryProto of data) {
const poseWorldLandmarksProto =
@ -452,7 +452,7 @@ export class PoseLandmarker extends VisionTaskRunner {
this.graphRunner.attachProtoVectorListener(
WORLD_LANDMARKS_STREAM, (binaryProto, timestamp) => {
this.adddJsWorldLandmarks(binaryProto);
this.addJsWorldLandmarks(binaryProto);
this.setLatestOutputTimestamp(timestamp);
});
this.graphRunner.attachEmptyPacketListener(

2
third_party/BUILD vendored
View File

@ -45,7 +45,7 @@ cc_library(
"@com_github_glog_glog//:glog",
],
"//mediapipe:windows": [
"@com_github_glog_glog//:glog",
"@com_github_glog_glog_windows//:glog",
],
"//conditions:default": [
"@com_github_glog_glog//:glog",

26
third_party/com_github_glog_glog_windows_patch.diff vendored Normal file
View File

@ -0,0 +1,26 @@
diff --git a/bazel/glog.bzl b/bazel/glog.bzl
index dacd934..d7b3d78 100644
--- a/bazel/glog.bzl
+++ b/bazel/glog.bzl
@@ -53,7 +53,6 @@ def glog_library(namespace = "google", with_gflags = 1, **kwargs):
)
common_copts = [
- "-std=c++14",
"-DGLOG_BAZEL_BUILD",
# Inject a C++ namespace.
"-DGOOGLE_NAMESPACE='%s'" % namespace,
@@ -145,7 +144,13 @@ def glog_library(namespace = "google", with_gflags = 1, **kwargs):
],
})
+ c14_opts = ["-std=c++14"]
+ c17_opts = ["-std=c++17"]
+
final_lib_copts = select({
+ "@bazel_tools//src/conditions:windows": c17_opts,
+ "//conditions:default": c14_opts,
+ }) + select({
"@bazel_tools//src/conditions:windows": common_copts + windows_only_copts,
"@bazel_tools//src/conditions:darwin": common_copts + linux_or_darwin_copts + darwin_only_copts,
"@bazel_tools//src/conditions:freebsd": common_copts + linux_or_darwin_copts + freebsd_only_copts,