mediapipe-nightly 0.10.21.post20241223__cp310-cp310-manylinux_2_28_x86_64.whl

mediapipe/util/sequence/media_sequence.py

@@ -0,0 +1,716 @@
+"""Copyright 2019 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.
+
+
+This script defines a large number of getters and setters for storing
+multimedia, such as video or audio, and related machine learning data in
+tf.train.SequenceExamples. These getters and setters simplify sharing
+data by enforcing common patterns for storing data in SequenceExample
+key-value pairs.
+
+The constants, macros, and functions are organized into 6 groups: clip
+metadata, clip label related, segment related, bounding-box related, image
+related, feature list related, and keyframe related. The following examples
+will walk through common task structures, but the relevant data to store can
+vary by task.
+
+The clip metadata group is generally data about the media and stored in the
+SequenceExample.context. Specifying the metadata enables media pipelines,
+such as MediaPipe, to retrieve that data. Typically, set_clip_data_path,
+set_clip_start_timestamp, and set_clip_end_timestamp define which data to use
+without storing the data itself. Example:
+  tensorflow.train.SequenceExample sequence
+  set_clip_data_path("/relative/path/to/data.mp4", sequence)
+  set_clip_start_timestamp(0, sequence)
+  set_clip_end_timestamp(10000000, sequence)  # 10 seconds in microseconds.
+
+The clip label group adds labels that apply to the entire media clip. To
+annotate that a video clip has a particular label, set the clip metadata
+above and also set the set_clip_label_index and set_clip_label_string. Most
+training pipelines will only use the label index or string, but we recommend
+storing both to improve readability while maintaining ease of use.
+Example:
+  set_clip_label_string(("run", "jump"), sequence)
+  set_Clip_label_index((35, 47), sequence)
+
+The segment group is generally data about time spans within the media clip
+and stored in the SequenceExample.context. In this code, continuous lengths
+of media are called clips, and each clip may have subregions of interest that
+are called segments. To annotate that a video clip has time spans with labels
+set the clip metadata above and use the functions set_segment_start_timestamp,
+set_segment_end_timestamp, set_segment_label_index, and
+set_segment_label_string. Most training pipelines will only use the label index
+or string, but we recommend storing both to improve readability while
+maintaining ease of use. By listing segments as times, the frame rate or other
+properties can change without affecting the labels.
+Example:
+  set_segment_start_timestamp((500000, 1000000), sequence)  # in microseconds
+  set_segment_end_timestamp((2000000, 6000000), sequence)
+  set_segment_label_index((35, 47), sequence)
+  set_segment_label_string(("run", "jump"), sequence)
+
+The bounding box group is useful for identifying spatio-temporal annotations
+for detection, tracking, or action recognition. The exact keys that are
+needed can vary by task, but to annotate a video clip for detection set the
+clip metadata above and use repeatedly call add_bbox, add_bbox_timestamp,
+add_bbox_label_index, and add_bbox_label_string. Most training pipelines will
+only use the label index or string, but we recommend storing both to improve
+readability while maintaining ease of use. Because bounding boxes are
+assigned to timepoints in a video, changing the image frame rate can can
+change the alignment. The media_sequence.h's ReconcileMetadata function can
+align bounding boxes to the nearest image.
+
+The image group is useful for storing data as sequential 2D arrays, typically
+encoded as bytes. Images can be RGB images stored as JPEG, discrete masks
+stored as PNG, or some other format. Parameters that are static over time are
+set in the context using set_image_width, set_image_height, set_image_format,
+etc. The series of frames and timestamps are then added with add_image_encoded
+and
+add_image_timestamp. For discrete masks, the class or instance indices can be
+mapped to labels or classes using
+set_class_segmentation_class_label_{index,string} and
+set_instance_segmentation_object_class_index.
+
+The feature list group is useful for storing audio and extracted features,
+such as per-frame embeddings. SequenceExamples only store lists of floats per
+timestep, so the dimensions are stored in the context to enable reshaping.
+For example, set_feature_dimensions and repeatedly calling add_feature_floats
+and add_feature_timestamp adds per-frame embeddings. The feature methods also
+support audio features.
+
+Macros for common patterns are created in media_sequence_util.py and are used
+here extensively. Because these macros are formulaic, I will only include a
+usage example here in the code rather than repeating documentation for every
+instance. This header defines additional functions to simplify working with
+MediaPipe types.
+
+Each msu.create_{TYPE}_context_feature takes a NAME and a KEY. It provides
+setters and getters for SequenceExamples and stores a single value under KEY
+in the context field. The provided functions are has_${NAME}, get_${NAME},
+set_${Name}, and clear_${NAME}.
+Eg.
+  tf.train.SequenceExample example
+  set_data_path("data_path", example)
+  if has_data_path(example):
+     data_path = get_data_path(example)
+     clear_data_path(example)
+
+Each msu.create_{TYPE}_list_context_feature takes a NAME and a KEY. It provides
+setters and getters for SequenceExamples and stores a sequence of values
+under KEY in the context field. The provided functions are has_${NAME},
+get_${NAME}, set_${Name}, clear_${NAME}, get_${NAME}_at, and add_${NAME}.
+Eg.
+  tf.train.SequenceExample example
+  set_clip_label_string(("run", "jump"), example)
+  if has_clip_label_string(example):
+     values = get_clip_label_string(example)
+     clear_clip_label_string(example)
+
+Each msu.create_{TYPE}_feature_list takes a NAME and a KEY. It provides setters
+and getters for SequenceExamples and stores a single value in each feature field
+under KEY of the feature_lists field. The provided functions are has_${NAME},
+get_${NAME}, clear_${NAME}, get_${NAME}_size, get_${NAME}_at, and add_${NAME}.
+  tf.train.SequenceExample example
+  add_image_timestamp(1000000, example)
+  add_image_timestamp(2000000, example)
+  if has_image_timestamp(example):
+    for i in range(get_image_timestamp_size()):
+      timestamp = get_image_timestamp_at(example, i)
+    clear_image_timestamp(example)
+
+Each VECTOR_{TYPE}_FEATURE_LIST takes a NAME and a KEY. It provides setters
+and getters for SequenceExamples and stores a sequence of values in each
+feature field under KEY of the feature_lists field. The provided functions
+are Has${NAME}, Get${NAME}, Clear${NAME}, Get${NAME}Size, Get${NAME}At, and
+Add${NAME}.
+  tf.train.SequenceExample example
+  add_bbox_label_string(("run", "jump"), example)
+  add_bbox_label_string(("run", "fall"), example)
+  if has_bbox_label_string(example):
+    for i in range(get_bbox_label_string_size(example)):
+      labels = get_bbox_label_string_at(example, i)
+    clear_bbox_label_string(example)
+
+As described in media_sequence_util.h, each of these functions can take an
+additional string prefix argument as their first argument. The prefix can
+be fixed with a new NAME by using functools.partial. Prefixes are used to
+identify common storage patterns (e.g. storing an image along with the height
+and width) under different names (e.g. storing a left and right image in a
+stereo pair.) An example creating functions such as
+add_left_image_encoded that adds a string under the key "LEFT/image/encoded"
+ add_left_image_encoded = msu.function_with_default(add_image_encoded, "LEFT")
+"""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+import numpy as np
+from mediapipe.util.sequence import media_sequence_util
+msu = media_sequence_util
+
+_HAS_DYNAMIC_ATTRIBUTES = True
+
+##################################  METADATA  #################################
+# A unique identifier for each example.
+EXAMPLE_ID_KEY = "example/id"
+# The name o fthe data set, including the version.
+EXAMPLE_DATASET_NAME_KEY = "example/dataset_name"
+# String flags or attributes for this example within a data set.
+EXAMPLE_DATASET_FLAG_STRING_KEY = "example/dataset/flag/string"
+# The relative path to the data on disk from some root directory.
+CLIP_DATA_PATH_KEY = "clip/data_path"
+# Any identifier for the media beyond the data path.
+CLIP_MEDIA_ID_KEY = "clip/media_id"
+# Yet another alternative identifier.
+ALTERNATIVE_CLIP_MEDIA_ID_KEY = "clip/alternative_media_id"
+# The encoded bytes for storing media directly in the SequenceExample.
+CLIP_ENCODED_MEDIA_BYTES_KEY = "clip/encoded_media_bytes"
+# The start time for the encoded media if not preserved during encoding.
+CLIP_ENCODED_MEDIA_START_TIMESTAMP_KEY = "clip/encoded_media_start_timestamp"
+# The start time, in microseconds, for the start of the clip in the media.
+CLIP_START_TIMESTAMP_KEY = "clip/start/timestamp"
+# The end time, in microseconds, for the end of the clip in the media.
+CLIP_END_TIMESTAMP_KEY = "clip/end/timestamp"
+# A list of label indices for this clip.
+CLIP_LABEL_INDEX_KEY = "clip/label/index"
+# A list of label strings for this clip.
+CLIP_LABEL_STRING_KEY = "clip/label/string"
+# A list of label confidences for this clip.
+CLIP_LABEL_CONFIDENCE_KEY = "clip/label/confidence"
+# A list of label start timestamps for this clip.
+CLIP_LABEL_START_TIMESTAMP_KEY = "clip/label/start/timestamp"
+# A list of label end timestamps for this clip.
+CLIP_LABEL_END_TIMESTAMP_KEY = "clip/label/end/timestamp"
+msu.create_bytes_context_feature(
+    "example_id", EXAMPLE_ID_KEY, module_dict=globals())
+msu.create_bytes_context_feature(
+    "example_dataset_name", EXAMPLE_DATASET_NAME_KEY, module_dict=globals())
+msu.create_bytes_list_context_feature(
+    "example_dataset_flag_string", EXAMPLE_DATASET_FLAG_STRING_KEY,
+    module_dict=globals())
+msu.create_bytes_context_feature(
+    "clip_media_id", CLIP_MEDIA_ID_KEY, module_dict=globals())
+msu.create_bytes_context_feature(
+    "clip_alternative_media_id", ALTERNATIVE_CLIP_MEDIA_ID_KEY,
+    module_dict=globals())
+msu.create_bytes_context_feature(
+    "clip_encoded_media_bytes", CLIP_ENCODED_MEDIA_BYTES_KEY,
+    module_dict=globals())
+msu.create_bytes_context_feature(
+    "clip_data_path", CLIP_DATA_PATH_KEY, module_dict=globals())
+msu.create_int_context_feature(
+    "clip_encoded_media_start_timestamp",
+    CLIP_ENCODED_MEDIA_START_TIMESTAMP_KEY, module_dict=globals())
+msu.create_int_context_feature(
+    "clip_start_timestamp", CLIP_START_TIMESTAMP_KEY, module_dict=globals())
+msu.create_int_context_feature(
+    "clip_end_timestamp", CLIP_END_TIMESTAMP_KEY, module_dict=globals())
+msu.create_bytes_list_context_feature(
+    "clip_label_string", CLIP_LABEL_STRING_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "clip_label_index", CLIP_LABEL_INDEX_KEY, module_dict=globals())
+msu.create_float_list_context_feature(
+    "clip_label_confidence", CLIP_LABEL_CONFIDENCE_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "clip_label_start_timestamp",
+    CLIP_LABEL_START_TIMESTAMP_KEY,
+    module_dict=globals())
+msu.create_int_list_context_feature(
+    "clip_label_end_timestamp",
+    CLIP_LABEL_END_TIMESTAMP_KEY,
+    module_dict=globals())
+
+##################################  SEGMENTS  #################################
+# A list of segment start times in microseconds.
+SEGMENT_START_TIMESTAMP_KEY = "segment/start/timestamp"
+# A list of indices marking the first frame index >= the start timestamp.
+SEGMENT_START_INDEX_KEY = "segment/start/index"
+# A list of segment end times in microseconds.
+SEGMENT_END_TIMESTAMP_KEY = "segment/end/timestamp"
+# A list of indices marking the last frame index <= the end timestamp.
+SEGMENT_END_INDEX_KEY = "segment/end/index"
+# A list with the label index for each segment.
+# Multiple labels for the same segment are encoded as repeated segments.
+SEGMENT_LABEL_INDEX_KEY = "segment/label/index"
+# A list with the label string for each segment.
+# Multiple labels for the same segment are encoded as repeated segments.
+SEGMENT_LABEL_STRING_KEY = "segment/label/string"
+# A list with the label confidence for each segment.
+# Multiple labels for the same segment are encoded as repeated segments.
+SEGMENT_LABEL_CONFIDENCE_KEY = "segment/label/confidence"
+msu.create_bytes_list_context_feature(
+    "segment_label_string", SEGMENT_LABEL_STRING_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "segment_start_timestamp",
+    SEGMENT_START_TIMESTAMP_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "segment_start_index", SEGMENT_START_INDEX_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "segment_end_timestamp", SEGMENT_END_TIMESTAMP_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "segment_end_index", SEGMENT_END_INDEX_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "segment_label_index", SEGMENT_LABEL_INDEX_KEY, module_dict=globals())
+msu.create_float_list_context_feature(
+    "segment_label_confidence",
+    SEGMENT_LABEL_CONFIDENCE_KEY, module_dict=globals())
+
+##########################  REGIONS / BOUNDING BOXES  #########################
+
+# Normalized coordinates of bounding boxes are provided in four lists to avoid
+# order ambiguity. We provide additional accessors for complete bounding boxes
+# below.
+REGION_BBOX_YMIN_KEY = "region/bbox/ymin"
+REGION_BBOX_XMIN_KEY = "region/bbox/xmin"
+REGION_BBOX_YMAX_KEY = "region/bbox/ymax"
+REGION_BBOX_XMAX_KEY = "region/bbox/xmax"
+# The point and radius can denote keypoints.
+REGION_POINT_X_KEY = "region/point/x"
+REGION_POINT_Y_KEY = "region/point/y"
+REGION_RADIUS_KEY = "region/radius"
+# The 3D point can denote keypoints.
+REGION_3D_POINT_X_KEY = "region/3d_point/x"
+REGION_3D_POINT_Y_KEY = "region/3d_point/y"
+REGION_3D_POINT_Z_KEY = "region/3d_point/z"
+# The number of regions at that timestep.
+REGION_NUM_REGIONS_KEY = "region/num_regions"
+# Whether that timestep is annotated for regions.
+# (Disambiguates between multiple meanings of num_regions = 0.)
+REGION_IS_ANNOTATED_KEY = "region/is_annotated"
+# A list indicating if each region is generated (1) or manually annotated (0)
+REGION_IS_GENERATED_KEY = "region/is_generated"
+# A list indicating if each region is occluded (1) or visible (0)
+REGION_IS_OCCLUDED_KEY = "region/is_occluded"
+# Lists with a label for each region.
+# Multiple labels for the same region require duplicating the region.
+REGION_LABEL_INDEX_KEY = "region/label/index"
+REGION_LABEL_STRING_KEY = "region/label/string"
+REGION_LABEL_CONFIDENCE_KEY = "region/label/confidence"
+# Lists with a track identifier for each region.
+# Multiple track identifier for the same region require duplicating the region.
+REGION_TRACK_INDEX_KEY = "region/track/index"
+REGION_TRACK_STRING_KEY = "region/track/string"
+REGION_TRACK_CONFIDENCE_KEY = "region/track/confidence"
+# Lists with a class for each region. In general, prefer to use the label
+# fields. These class fields exist to distinguish tracks when different classes
+# have overlapping track ids.
+REGION_CLASS_INDEX_KEY = "region/class/index"
+REGION_CLASS_STRING_KEY = "region/class/string"
+REGION_CLASS_CONFIDENCE_KEY = "region/class/confidence"
+# The timestamp of the region annotation in microseconds.
+REGION_TIMESTAMP_KEY = "region/timestamp"
+# The original timestamp in microseconds for region annotations.
+# If regions are aligned to image frames, this field preserves the original
+# timestamps.
+REGION_UNMODIFIED_TIMESTAMP_KEY = "region/unmodified_timestamp"
+# The list of region parts expected in this example.
+REGION_PARTS_KEY = "region/parts"
+# The dimensions of each embedding per region / bounding box.
+REGION_EMBEDDING_DIMENSIONS_PER_REGION_KEY = (
+    "region/embedding/dimensions_per_region")
+# The format encoding embeddings as strings.
+REGION_EMBEDDING_FORMAT_KEY = "region/embedding/format"
+# An embedding for each region. The length of each list must be the product of
+# the number of regions and the product of the embedding dimensions.
+REGION_EMBEDDING_FLOAT_KEY = "region/embedding/float"
+# A string encoded embedding for each regions.
+REGION_EMBEDDING_ENCODED_KEY = "region/embedding/encoded"
+# The confidence of the embedding.
+REGION_EMBEDDING_CONFIDENCE_KEY = "region/embedding/confidence"
+
+
+def _create_region_with_prefix(name, prefix):
+  """Create multiple accessors for region based data."""
+  msu.create_int_feature_list(name + "_num_regions", REGION_NUM_REGIONS_KEY,
+                              prefix=prefix, module_dict=globals())
+  msu.create_int_feature_list(name + "_is_annotated", REGION_IS_ANNOTATED_KEY,
+                              prefix=prefix, module_dict=globals())
+  msu.create_int_list_feature_list(
+      name + "_is_occluded", REGION_IS_OCCLUDED_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_list_feature_list(
+      name + "_is_generated", REGION_IS_GENERATED_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_feature_list(name + "_timestamp", REGION_TIMESTAMP_KEY,
+                              prefix=prefix, module_dict=globals())
+  msu.create_int_feature_list(
+      name + "_unmodified_timestamp", REGION_UNMODIFIED_TIMESTAMP_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_feature_list(
+      name + "_label_string", REGION_LABEL_STRING_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_list_feature_list(
+      name + "_label_index", REGION_LABEL_INDEX_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(
+      name + "_label_confidence", REGION_LABEL_CONFIDENCE_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_feature_list(
+      name + "_class_string", REGION_CLASS_STRING_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_list_feature_list(
+      name + "_class_index", REGION_CLASS_INDEX_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(
+      name + "_class_confidence", REGION_CLASS_CONFIDENCE_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_feature_list(
+      name + "_track_string", REGION_TRACK_STRING_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_list_feature_list(
+      name + "_track_index", REGION_TRACK_INDEX_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(
+      name + "_track_confidence", REGION_TRACK_CONFIDENCE_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_ymin", REGION_BBOX_YMIN_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_xmin", REGION_BBOX_XMIN_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_ymax", REGION_BBOX_YMAX_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_xmax", REGION_BBOX_XMAX_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_point_x", REGION_POINT_X_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_point_y", REGION_POINT_Y_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(
+      name + "_3d_point_x", REGION_3D_POINT_X_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(
+      name + "_3d_point_y", REGION_3D_POINT_Y_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(
+      name + "_3d_point_z", REGION_3D_POINT_Z_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_context_feature(name + "_parts",
+                                        REGION_PARTS_KEY,
+                                        prefix=prefix, module_dict=globals())
+  msu.create_float_list_context_feature(
+      name + "_embedding_dimensions_per_region",
+      REGION_EMBEDDING_DIMENSIONS_PER_REGION_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_bytes_context_feature(name + "_embedding_format",
+                                   REGION_EMBEDDING_FORMAT_KEY,
+                                   prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_embedding_floats",
+                                     REGION_EMBEDDING_FLOAT_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_feature_list(name + "_embedding_encoded",
+                                     REGION_EMBEDDING_ENCODED_KEY,
+                                     prefix=prefix, module_dict=globals())
+  msu.create_float_list_feature_list(name + "_embedding_confidence",
+                                     REGION_EMBEDDING_CONFIDENCE_KEY,
+                                     prefix=prefix, module_dict=globals())
+  # pylint: disable=undefined-variable
+  def get_prefixed_bbox_at(index, sequence_example, prefix):
+    return np.stack((
+        get_bbox_ymin_at(index, sequence_example, prefix=prefix),
+        get_bbox_xmin_at(index, sequence_example, prefix=prefix),
+        get_bbox_ymax_at(index, sequence_example, prefix=prefix),
+        get_bbox_xmax_at(index, sequence_example, prefix=prefix)),
+                    1)
+  def add_prefixed_bbox(values, sequence_example, prefix):
+    values = np.array(values)
+    if values.size == 0:
+      add_bbox_ymin([], sequence_example, prefix=prefix)
+      add_bbox_xmin([], sequence_example, prefix=prefix)
+      add_bbox_ymax([], sequence_example, prefix=prefix)
+      add_bbox_xmax([], sequence_example, prefix=prefix)
+    else:
+      add_bbox_ymin(values[:, 0], sequence_example, prefix=prefix)
+      add_bbox_xmin(values[:, 1], sequence_example, prefix=prefix)
+      add_bbox_ymax(values[:, 2], sequence_example, prefix=prefix)
+      add_bbox_xmax(values[:, 3], sequence_example, prefix=prefix)
+  def get_prefixed_bbox_size(sequence_example, prefix):
+    return get_bbox_ymin_size(sequence_example, prefix=prefix)
+  def has_prefixed_bbox(sequence_example, prefix):
+    return has_bbox_ymin(sequence_example, prefix=prefix)
+  def clear_prefixed_bbox(sequence_example, prefix):
+    clear_bbox_ymin(sequence_example, prefix=prefix)
+    clear_bbox_xmin(sequence_example, prefix=prefix)
+    clear_bbox_ymax(sequence_example, prefix=prefix)
+    clear_bbox_xmax(sequence_example, prefix=prefix)
+  def get_prefixed_point_at(index, sequence_example, prefix):
+    return np.stack((
+        get_bbox_point_y_at(index, sequence_example, prefix=prefix),
+        get_bbox_point_x_at(index, sequence_example, prefix=prefix)),
+                    1)
+  def add_prefixed_point(values, sequence_example, prefix):
+    add_bbox_point_y(values[:, 0], sequence_example, prefix=prefix)
+    add_bbox_point_x(values[:, 1], sequence_example, prefix=prefix)
+  def get_prefixed_point_size(sequence_example, prefix):
+    return get_bbox_point_y_size(sequence_example, prefix=prefix)
+  def has_prefixed_point(sequence_example, prefix):
+    return has_bbox_point_y(sequence_example, prefix=prefix)
+  def clear_prefixed_point(sequence_example, prefix):
+    clear_bbox_point_y(sequence_example, prefix=prefix)
+    clear_bbox_point_x(sequence_example, prefix=prefix)
+  def get_prefixed_3d_point_at(index, sequence_example, prefix):
+    return np.stack((
+        get_bbox_3d_point_x_at(index, sequence_example, prefix=prefix),
+        get_bbox_3d_point_y_at(index, sequence_example, prefix=prefix),
+        get_bbox_3d_point_z_at(index, sequence_example, prefix=prefix)),
+                    1)
+  def add_prefixed_3d_point(values, sequence_example, prefix):
+    add_bbox_3d_point_x(values[:, 0], sequence_example, prefix=prefix)
+    add_bbox_3d_point_y(values[:, 1], sequence_example, prefix=prefix)
+    add_bbox_3d_point_z(values[:, 2], sequence_example, prefix=prefix)
+  def get_prefixed_3d_point_size(sequence_example, prefix):
+    return get_bbox_3d_point_x_size(sequence_example, prefix=prefix)
+  def has_prefixed_3d_point(sequence_example, prefix):
+    return has_bbox_3d_point_x(sequence_example, prefix=prefix)
+  def clear_prefixed_3d_point(sequence_example, prefix):
+    clear_bbox_3d_point_x(sequence_example, prefix=prefix)
+    clear_bbox_3d_point_y(sequence_example, prefix=prefix)
+    clear_bbox_3d_point_z(sequence_example, prefix=prefix)
+  # pylint: enable=undefined-variable
+  msu.add_functions_to_module({
+      "get_" + name + "_at":
+          msu.function_with_default(get_prefixed_bbox_at, prefix),
+      "add_" + name:
+          msu.function_with_default(add_prefixed_bbox, prefix),
+      "get_" + name + "_size":
+          msu.function_with_default(get_prefixed_bbox_size, prefix),
+      "has_" + name:
+          msu.function_with_default(has_prefixed_bbox, prefix),
+      "clear_" + name:
+          msu.function_with_default(clear_prefixed_bbox, prefix),
+  }, module_dict=globals())
+  msu.add_functions_to_module({
+      "get_" + name + "_point_at":
+          msu.function_with_default(get_prefixed_point_at, prefix),
+      "add_" + name + "_point":
+          msu.function_with_default(add_prefixed_point, prefix),
+      "get_" + name + "_point_size":
+          msu.function_with_default(get_prefixed_point_size, prefix),
+      "has_" + name + "_point":
+          msu.function_with_default(has_prefixed_point, prefix),
+      "clear_" + name + "_point":
+          msu.function_with_default(clear_prefixed_point, prefix),
+  }, module_dict=globals())
+  msu.add_functions_to_module({
+      "get_" + name + "_3d_point_at":
+          msu.function_with_default(get_prefixed_3d_point_at, prefix),
+      "add_" + name + "_3d_point":
+          msu.function_with_default(add_prefixed_3d_point, prefix),
+      "get_" + name + "_3d_point_size":
+          msu.function_with_default(get_prefixed_3d_point_size, prefix),
+      "has_" + name + "_3d_point":
+          msu.function_with_default(has_prefixed_3d_point, prefix),
+      "clear_" + name + "_3d_point":
+          msu.function_with_default(clear_prefixed_3d_point, prefix),
+  }, module_dict=globals())
+
+
+PREDICTED_PREFIX = "PREDICTED"
+_create_region_with_prefix("bbox", "")
+_create_region_with_prefix("predicted_bbox", PREDICTED_PREFIX)
+
+
+###################################  IMAGES  #################################
+# The format the images are encoded as (e.g. "JPEG", "PNG")
+IMAGE_FORMAT_KEY = "image/format"
+# The number of channels in the image.
+IMAGE_CHANNELS_KEY = "image/channels"
+# The colorspace of the iamge.
+IMAGE_COLORSPACE_KEY = "image/colorspace"
+# The height of the image in pixels.
+IMAGE_HEIGHT_KEY = "image/height"
+# The width of the image in pixels.
+IMAGE_WIDTH_KEY = "image/width"
+# frame rate in images/second of media.
+IMAGE_FRAME_RATE_KEY = "image/frame_rate"
+# The maximum values if the images were saturated and normalized for encoding.
+IMAGE_SATURATION_KEY = "image/saturation"
+# The listing from discrete image values (as indices) to class indices.
+IMAGE_CLASS_LABEL_INDEX_KEY = "image/class/label/index"
+# The listing from discrete image values (as indices) to class strings.
+IMAGE_CLASS_LABEL_STRING_KEY = "image/class/label/string"
+# The listing from discrete instance indices to class indices they embody.
+IMAGE_OBJECT_CLASS_INDEX_KEY = "image/object/class/index"
+# The encoded image frame.
+IMAGE_ENCODED_KEY = "image/encoded"
+# Multiple images from the same timestep (e.g. multiview video).
+IMAGE_MULTI_ENCODED_KEY = "image/multi_encoded"
+# The timestamp of the frame in microseconds.
+IMAGE_TIMESTAMP_KEY = "image/timestamp"
+# A per image label if specific frames have labels.
+# If time spans have labels, segments are preferred to allow changing rates.
+IMAGE_LABEL_INDEX_KEY = "image/label/index"
+IMAGE_LABEL_STRING_KEY = "image/label/string"
+IMAGE_LABEL_CONFIDENCE_KEY = "image/label/confidence"
+# The path of the image file if it did not come from a media clip.
+IMAGE_DATA_PATH_KEY = "image/data_path"
+
+
+def _create_image_with_prefix(name, prefix):
+  """Create multiple accessors for image based data."""
+  msu.create_bytes_context_feature(name + "_format", IMAGE_FORMAT_KEY,
+                                   prefix=prefix, module_dict=globals())
+  msu.create_bytes_context_feature(name + "_colorspace", IMAGE_COLORSPACE_KEY,
+                                   prefix=prefix, module_dict=globals())
+  msu.create_int_context_feature(name + "_channels", IMAGE_CHANNELS_KEY,
+                                 prefix=prefix, module_dict=globals())
+  msu.create_int_context_feature(name + "_height", IMAGE_HEIGHT_KEY,
+                                 prefix=prefix, module_dict=globals())
+  msu.create_int_context_feature(name + "_width", IMAGE_WIDTH_KEY,
+                                 prefix=prefix, module_dict=globals())
+  msu.create_bytes_feature_list(name + "_encoded", IMAGE_ENCODED_KEY,
+                                prefix=prefix, module_dict=globals())
+  msu.create_float_context_feature(name + "_frame_rate", IMAGE_FRAME_RATE_KEY,
+                                   prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_context_feature(
+      name + "_class_label_string", IMAGE_CLASS_LABEL_STRING_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_list_context_feature(
+      name + "_class_label_index", IMAGE_CLASS_LABEL_INDEX_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_int_list_context_feature(
+      name + "_object_class_index", IMAGE_OBJECT_CLASS_INDEX_KEY,
+      prefix=prefix, module_dict=globals())
+  msu.create_bytes_context_feature(name + "_data_path", IMAGE_DATA_PATH_KEY,
+                                   prefix=prefix, module_dict=globals())
+  msu.create_int_feature_list(name + "_timestamp", IMAGE_TIMESTAMP_KEY,
+                              prefix=prefix, module_dict=globals())
+  msu.create_bytes_list_feature_list(name + "_multi_encoded",
+                                     IMAGE_MULTI_ENCODED_KEY, prefix=prefix,
+                                     module_dict=globals())
+FORWARD_FLOW_PREFIX = "FORWARD_FLOW"
+CLASS_SEGMENTATION_PREFIX = "CLASS_SEGMENTATION"
+INSTANCE_SEGMENTATION_PREFIX = "INSTANCE_SEGMENTATION"
+_create_image_with_prefix("image", "")
+_create_image_with_prefix("forward_flow", FORWARD_FLOW_PREFIX)
+_create_image_with_prefix("class_segmentation", CLASS_SEGMENTATION_PREFIX)
+_create_image_with_prefix("instance_segmentation", INSTANCE_SEGMENTATION_PREFIX)
+
+##################################  TEXT  #################################
+# Which language text tokens are likely to be in.
+TEXT_LANGUAGE_KEY = "text/language"
+# A large block of text that applies to the media.
+TEXT_CONTEXT_CONTENT_KEY = "text/context/content"
+# A large block of text that applies to the media as token ids.
+TEXT_CONTEXT_TOKEN_ID_KEY = "text/context/token_id"
+# A large block of text that applies to the media as embeddings.
+TEXT_CONTEXT_EMBEDDING_KEY = "text/context/embedding"
+
+# The text contents for a given time.
+TEXT_CONTENT_KEY = "text/content"
+# The start time for the text becoming relevant.
+TEXT_TIMESTAMP_KEY = "text/timestamp"
+# The duration where the text is relevant.
+TEXT_DURATION_KEY = "text/duration"
+# The confidence that this is the correct text.
+TEXT_CONFIDENCE_KEY = "text/confidence"
+# A floating point embedding corresponding to the text.
+TEXT_EMBEDDING_KEY = "text/embedding"
+# An integer id corresponding to the text.
+TEXT_TOKEN_ID_KEY = "text/token/id"
+
+msu.create_bytes_context_feature(
+    "text_language", TEXT_LANGUAGE_KEY, module_dict=globals())
+msu.create_bytes_context_feature(
+    "text_context_content", TEXT_CONTEXT_CONTENT_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "text_context_token_id", TEXT_CONTEXT_TOKEN_ID_KEY, module_dict=globals())
+msu.create_float_list_context_feature(
+    "text_context_embedding", TEXT_CONTEXT_EMBEDDING_KEY, module_dict=globals())
+msu.create_bytes_feature_list(
+    "text_content", TEXT_CONTENT_KEY, module_dict=globals())
+msu.create_int_feature_list(
+    "text_timestamp", TEXT_TIMESTAMP_KEY, module_dict=globals())
+msu.create_int_feature_list(
+    "text_duration", TEXT_DURATION_KEY, module_dict=globals())
+msu.create_float_feature_list(
+    "text_confidence", TEXT_CONFIDENCE_KEY, module_dict=globals())
+msu.create_float_list_feature_list(
+    "text_embedding", TEXT_EMBEDDING_KEY, module_dict=globals())
+msu.create_int_feature_list(
+    "text_token_id", TEXT_TOKEN_ID_KEY, module_dict=globals())
+
+##################################  FEATURES  #################################
+# The dimensions of the feature.
+FEATURE_DIMENSIONS_KEY = "feature/dimensions"
+# The rate the features are extracted per second of media.
+FEATURE_RATE_KEY = "feature/rate"
+# The encoding format if any for the feature.
+FEATURE_BYTES_FORMAT_KEY = "feature/bytes/format"
+# For audio, the rate the samples are extracted per second of media.
+FEATURE_SAMPLE_RATE_KEY = "feature/sample_rate"
+# For audio, the number of channels per extracted feature.
+FEATURE_NUM_CHANNELS_KEY = "feature/num_channels"
+# For audio, th enumber of samples per extracted feature.
+FEATURE_NUM_SAMPLES_KEY = "feature/num_samples"
+# For audio, the rate the features are extracted per second of media.
+FEATURE_PACKET_RATE_KEY = "feature/packet_rate"
+# For audio, the original audio sampling rate the feature is derived from.
+FEATURE_AUDIO_SAMPLE_RATE_KEY = "feature/audio_sample_rate"
+# The feature as a list of floats.
+FEATURE_FLOATS_KEY = "feature/floats"
+# The feature as a list of bytes. May be encoded.
+FEATURE_BYTES_KEY = "feature/bytes"
+# The feature as a list of ints.
+FEATURE_INTS_KEY = "feature/ints"
+# The timestamp, in microseconds, of the feature.
+FEATURE_TIMESTAMP_KEY = "feature/timestamp"
+# It is occasionally useful to indicate that a feature applies to a given range.
+# This should be used for features only and annotations should be provided as
+# segments.
+FEATURE_DURATION_KEY = "feature/duration"
+# Encodes an optional confidence score for the generated features.
+FEATURE_CONFIDENCE_KEY = "feature/confidence"
+# The feature as a list of floats in the context.
+CONTEXT_FEATURE_FLOATS_KEY = "context_feature/floats"
+# The feature as a list of bytes in the context. May be encoded.
+CONTEXT_FEATURE_BYTES_KEY = "context_feature/bytes"
+# The feature as a list of ints in the context.
+CONTEXT_FEATURE_INTS_KEY = "context_feature/ints"
+
+msu.create_int_list_context_feature(
+    "feature_dimensions", FEATURE_DIMENSIONS_KEY, module_dict=globals())
+msu.create_float_context_feature(
+    "feature_rate", FEATURE_RATE_KEY, module_dict=globals())
+msu.create_bytes_context_feature(
+    "feature_bytes_format", FEATURE_BYTES_FORMAT_KEY, module_dict=globals())
+msu.create_float_context_feature(
+    "feature_sample_rate", FEATURE_SAMPLE_RATE_KEY, module_dict=globals())
+msu.create_int_context_feature(
+    "feature_num_channels", FEATURE_NUM_CHANNELS_KEY, module_dict=globals())
+msu.create_int_context_feature(
+    "feature_num_samples", FEATURE_NUM_SAMPLES_KEY, module_dict=globals())
+msu.create_float_context_feature(
+    "feature_packet_rate", FEATURE_PACKET_RATE_KEY, module_dict=globals())
+msu.create_float_context_feature(
+    "feature_audio_sample_rate", FEATURE_AUDIO_SAMPLE_RATE_KEY,
+    module_dict=globals())
+msu.create_float_list_feature_list(
+    "feature_floats", FEATURE_FLOATS_KEY, module_dict=globals())
+msu.create_bytes_list_feature_list(
+    "feature_bytes", FEATURE_BYTES_KEY, module_dict=globals())
+msu.create_int_list_feature_list(
+    "feature_ints", FEATURE_INTS_KEY, module_dict=globals())
+msu.create_int_feature_list(
+    "feature_timestamp", FEATURE_TIMESTAMP_KEY, module_dict=globals())
+msu.create_int_list_feature_list(
+    "feature_duration", FEATURE_DURATION_KEY, module_dict=globals())
+msu.create_float_list_feature_list(
+    "feature_confidence", FEATURE_CONFIDENCE_KEY, module_dict=globals())
+msu.create_float_list_context_feature(
+    "context_feature_floats", CONTEXT_FEATURE_FLOATS_KEY, module_dict=globals())
+msu.create_bytes_list_context_feature(
+    "context_feature_bytes", CONTEXT_FEATURE_BYTES_KEY, module_dict=globals())
+msu.create_int_list_context_feature(
+    "context_feature_ints", CONTEXT_FEATURE_INTS_KEY, module_dict=globals())
+