mediapipe/docs/framework_concepts/graphs.md
MediaPipe Team 48aa88f39d Change object detector learning rate decay to cosine decay.
PiperOrigin-RevId: 527337105
2023-04-26 12:13:17 -07:00

400 lines
13 KiB
Markdown
Raw Permalink Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
layout: forward
target: https://developers.google.com/mediapipe/framework/framework_concepts/graphs
title: Graphs
parent: Framework Concepts
nav_order: 2
---
# Graphs
{: .no_toc }
1. TOC
{:toc}
---
**Attention:** *Thanks for your interest in MediaPipe! We have moved to
[https://developers.google.com/mediapipe](https://developers.google.com/mediapipe)
as the primary developer documentation site for MediaPipe as of April 3, 2023.*
----
## Graph
A `CalculatorGraphConfig` proto specifies the topology and functionality of a
MediaPipe graph. Each `node` in the graph represents a particular calculator or
subgraph, and specifies necessary configurations, such as registered
calculator/subgraph type, inputs, outputs and optional fields, such as
node-specific options, input policy and executor, discussed in
[Synchronization](synchronization.md).
`CalculatorGraphConfig` has several other fields to configure global graph-level
settings, e.g. graph executor configs, number of threads, and maximum queue size
of input streams. Several graph-level settings are useful for tuning the
performance of the graph on different platforms (e.g., desktop v.s. mobile). For
instance, on mobile, attaching a heavy model-inference calculator to a separate
executor can improve the performance of a real-time application since this
enables thread locality.
Below is a trivial `CalculatorGraphConfig` example where we have series of
passthrough calculators :
```proto
# This graph named main_pass_throughcals_nosubgraph.pbtxt contains 4
# passthrough calculators.
input_stream: "in"
output_stream: "out"
node {
calculator: "PassThroughCalculator"
input_stream: "in"
output_stream: "out1"
}
node {
calculator: "PassThroughCalculator"
input_stream: "out1"
output_stream: "out2"
}
node {
calculator: "PassThroughCalculator"
input_stream: "out2"
output_stream: "out3"
}
node {
calculator: "PassThroughCalculator"
input_stream: "out3"
output_stream: "out"
}
```
MediaPipe offers an alternative `C++` representation for complex graphs (e.g. ML pipelines, handling model metadata, optional nodes, etc.). The above graph may look like:
```c++
CalculatorGraphConfig BuildGraphConfig() {
Graph graph;
// Graph inputs
Stream<AnyType> in = graph.In(0).SetName("in");
auto pass_through_fn = [](Stream<AnyType> in,
Graph& graph) -> Stream<AnyType> {
auto& node = graph.AddNode("PassThroughCalculator");
in.ConnectTo(node.In(0));
return node.Out(0);
};
Stream<AnyType> out1 = pass_through_fn(in, graph);
Stream<AnyType> out2 = pass_through_fn(out1, graph);
Stream<AnyType> out3 = pass_through_fn(out2, graph);
Stream<AnyType> out4 = pass_through_fn(out3, graph);
// Graph outputs
out4.SetName("out").ConnectTo(graph.Out(0));
return graph.GetConfig();
}
```
See more details in [Building Graphs in C++](building_graphs_cpp.md)
## Subgraph
To modularize a `CalculatorGraphConfig` into sub-modules and assist with re-use
of perception solutions, a MediaPipe graph can be defined as a `Subgraph`. The
public interface of a subgraph consists of a set of input and output streams
similar to a calculator's public interface. The subgraph can then be included in
a `CalculatorGraphConfig` as if it were a calculator. When a MediaPipe graph is
loaded from a `CalculatorGraphConfig`, each subgraph node is replaced by the
corresponding graph of calculators. As a result, the semantics and performance
of the subgraph is identical to the corresponding graph of calculators.
Below is an example of how to create a subgraph named `TwoPassThroughSubgraph`.
1. Defining the subgraph.
```proto
# This subgraph is defined in two_pass_through_subgraph.pbtxt
# and is registered as "TwoPassThroughSubgraph"
type: "TwoPassThroughSubgraph"
input_stream: "out1"
output_stream: "out3"
node {
calculator: "PassThroughCalculator"
input_stream: "out1"
output_stream: "out2"
}
node {
calculator: "PassThroughCalculator"
input_stream: "out2"
output_stream: "out3"
}
```
The public interface to the subgraph consists of:
* Graph input streams
* Graph output streams
* Graph input side packets
* Graph output side packets
2. Register the subgraph using BUILD rule `mediapipe_simple_subgraph`. The
parameter `register_as` defines the component name for the new subgraph.
```proto
# Small section of BUILD file for registering the "TwoPassThroughSubgraph"
# subgraph for use by main graph main_pass_throughcals.pbtxt
mediapipe_simple_subgraph(
name = "twopassthrough_subgraph",
graph = "twopassthrough_subgraph.pbtxt",
register_as = "TwoPassThroughSubgraph",
deps = [
"//mediapipe/calculators/core:pass_through_calculator",
"//mediapipe/framework:calculator_graph",
],
)
```
3. Use the subgraph in the main graph.
```proto
# This main graph is defined in main_pass_throughcals.pbtxt
# using subgraph called "TwoPassThroughSubgraph"
input_stream: "in"
node {
calculator: "PassThroughCalculator"
input_stream: "in"
output_stream: "out1"
}
node {
calculator: "TwoPassThroughSubgraph"
input_stream: "out1"
output_stream: "out3"
}
node {
calculator: "PassThroughCalculator"
input_stream: "out3"
output_stream: "out4"
}
```
## Graph Options
It is possible to specify a "graph options" protobuf for a MediaPipe graph
similar to the [`Calculator Options`](calculators.md#calculator-options)
protobuf specified for a MediaPipe calculator. These "graph options" can be
specified where a graph is invoked, and used to populate calculator options and
subgraph options within the graph.
In a `CalculatorGraphConfig`, graph options can be specified for a subgraph
exactly like calculator options, as shown below:
```
node {
calculator: "FlowLimiterCalculator"
input_stream: "image"
output_stream: "throttled_image"
node_options: {
[type.googleapis.com/mediapipe.FlowLimiterCalculatorOptions] {
max_in_flight: 1
}
}
}
node {
calculator: "FaceDetectionSubgraph"
input_stream: "IMAGE:throttled_image"
node_options: {
[type.googleapis.com/mediapipe.FaceDetectionOptions] {
tensor_width: 192
tensor_height: 192
}
}
}
```
In a `CalculatorGraphConfig`, graph options can be accepted and used to populate
calculator options, as shown below:
```
graph_options: {
[type.googleapis.com/mediapipe.FaceDetectionOptions] {}
}
node: {
calculator: "ImageToTensorCalculator"
input_stream: "IMAGE:image"
node_options: {
[type.googleapis.com/mediapipe.ImageToTensorCalculatorOptions] {
keep_aspect_ratio: true
border_mode: BORDER_ZERO
}
}
option_value: "output_tensor_width:options/tensor_width"
option_value: "output_tensor_height:options/tensor_height"
}
node {
calculator: "InferenceCalculator"
node_options: {
[type.googleapis.com/mediapipe.InferenceCalculatorOptions] {}
}
option_value: "delegate:options/delegate"
option_value: "model_path:options/model_path"
}
```
In this example, the `FaceDetectionSubgraph` accepts graph option protobuf
`FaceDetectionOptions`. The `FaceDetectionOptions` is used to define some field
values in the calculator options `ImageToTensorCalculatorOptions` and some field
values in the subgraph options `InferenceCalculatorOptions`. The field values
are defined using the `option_value:` syntax.
In the `CalculatorGraphConfig::Node` protobuf, the fields `node_options:` and
`option_value:` together define the option values for a calculator such as
`ImageToTensorCalculator`. The `node_options:` field defines a set of literal
constant values using the text protobuf syntax. Each `option_value:` field
defines the value for one protobuf field using information from the enclosing
graph, specifically from field values of the graph options of the enclosing
graph. In the example above, the `option_value:`
`"output_tensor_width:options/tensor_width"` defines the field
`ImageToTensorCalculatorOptions.output_tensor_width` using the value of
`FaceDetectionOptions.tensor_width`.
The syntax of `option_value:` is similar to the syntax of `input_stream:`. The
syntax is `option_value: "LHS:RHS"`. The LHS identifies a calculator option
field and the RHS identifies a graph option field. More specifically, the LHS
and RHS each consists of a series of protobuf field names identifying nested
protobuf messages and fields separated by '/'. This is known as the "ProtoPath"
syntax. Nested messages that are referenced in the LHS or RHS must already be
defined in the enclosing protobuf in order to be traversed using
`option_value:`.
## Cycles
<!-- TODO: add discussion of PreviousLoopbackCalculator -->
By default, MediaPipe requires calculator graphs to be acyclic and treats cycles
in a graph as errors. If a graph is intended to have cycles, the cycles need to
be annotated in the graph config. This page describes how to do that.
NOTE: The current approach is experimental and subject to change. We welcome
your feedback.
Please use the `CalculatorGraphTest.Cycle` unit test in
`mediapipe/framework/calculator_graph_test.cc` as sample code. Shown below is
the cyclic graph in the test. The `sum` output of the adder is the sum of the
integers generated by the integer source calculator.
![a cyclic graph that adds a stream of integers](https://mediapipe.dev/images/cyclic_integer_sum_graph.svg "A cyclic graph")
This simple graph illustrates all the issues in supporting cyclic graphs.
### Back Edge Annotation
We require that an edge in each cycle be annotated as a back edge. This allows
MediaPipes topological sort to work, after removing all the back edges.
There are usually multiple ways to select the back edges. Which edges are marked
as back edges affects which nodes are considered as upstream and which nodes are
considered as downstream, which in turn affects the priorities MediaPipe assigns
to the nodes.
For example, the `CalculatorGraphTest.Cycle` test marks the `old_sum` edge as a
back edge, so the Delay node is considered as a downstream node of the adder
node and is given a higher priority. Alternatively, we could mark the `sum`
input to the delay node as the back edge, in which case the delay node would be
considered as an upstream node of the adder node and is given a lower priority.
### Initial Packet
For the adder calculator to be runnable when the first integer from the integer
source arrives, we need an initial packet, with value 0 and with the same
timestamp, on the `old_sum` input stream to the adder. This initial packet
should be output by the delay calculator in the `Open()` method.
### Delay in a Loop
Each loop should incur a delay to align the previous `sum` output with the next
integer input. This is also done by the delay node. So the delay node needs to
know the following about the timestamps of the integer source calculator:
* The timestamp of the first output.
* The timestamp delta between successive outputs.
We plan to add an alternative scheduling policy that only cares about packet
ordering and ignores packet timestamps, which will eliminate this inconvenience.
### Early Termination of a Calculator When One Input Stream is Done
By default, MediaPipe calls the `Close()` method of a non-source calculator when
all of its input streams are done. In the example graph, we want to stop the
adder node as soon as the integer source is done. This is accomplished by
configuring the adder node with an alternative input stream handler,
`EarlyCloseInputStreamHandler`.
### Relevant Source Code
#### Delay Calculator
Note the code in `Open()` that outputs the initial packet and the code in
`Process()` that adds a (unit) delay to input packets. As noted above, this
delay node assumes that its output stream is used alongside an input stream with
packet timestamps 0, 1, 2, 3, ...
```c++
class UnitDelayCalculator : public Calculator {
public:
 static absl::Status FillExpectations(
     const CalculatorOptions& extendable_options, PacketTypeSet* inputs,
     PacketTypeSet* outputs, PacketTypeSet* input_side_packets) {
   inputs->Index(0)->Set<int>("An integer.");
   outputs->Index(0)->Set<int>("The input delayed by one time unit.");
   return absl::OkStatus();
 }
 absl::Status Open() final {
   Output()->Add(new int(0), Timestamp(0));
   return absl::OkStatus();
 }
 absl::Status Process() final {
   const Packet& packet = Input()->Value();
   Output()->AddPacket(packet.At(packet.Timestamp().NextAllowedInStream()));
   return absl::OkStatus();
 }
};
```
#### Graph Config
Note the `back_edge` annotation and the alternative `input_stream_handler`.
```proto
node {
  calculator: 'GlobalCountSourceCalculator'
  input_side_packet: 'global_counter'
  output_stream: 'integers'
}
node {
  calculator: 'IntAdderCalculator'
  input_stream: 'integers'
  input_stream: 'old_sum'
  input_stream_info: {
    tag_index: ':1' # 'old_sum'
    back_edge: true
  }
  output_stream: 'sum'
  input_stream_handler {
    input_stream_handler: 'EarlyCloseInputStreamHandler'
  }
}
node {
  calculator: 'UnitDelayCalculator'
  input_stream: 'sum'
  output_stream: 'old_sum'
}
```