Python Model API package: add main documentation

ci/check-documentation.py

@@ -78,6 +78,11 @@ def complain(message):
     omz_github_url = 'https://github.com/openvinotoolkit/open_model_zoo/'
 
     for md_path in sorted(all_md_files):
+
+        # skip the checks for Python Model API package
+        if 'model_api' in str(md_path):
+            continue
+
         referenced_md_files = set()
 
         md_path_rel = md_path.relative_to(OMZ_ROOT)

ci/prepare-documentation.py

@@ -382,6 +382,11 @@ def main():
         title='OMZ Model API OVMS adapter')
     ovms_adapter_element.attrib[XML_ID_ATTRIBUTE] = 'omz_model_api_ovms_adapter'
 
+    model_api_element = add_page(output_root, navindex_element, id='omz_python_model_api',
+        path='demos/common/python/openvino/model_zoo/model_api/README.md',
+        title='OMZ Python Model API')
+    model_api_element.attrib[XML_ID_ATTRIBUTE] = 'omz_python_model_api'
+
     for md_path in all_md_paths:
         if md_path not in documentation_md_paths:
             raise RuntimeError(f'{all_md_paths[md_path]}: '

demos/README.md

@@ -195,35 +195,7 @@ cmake -A x64 <open_model_zoo>/demos
 
 ### <a name="model_api_installation"></a>Python\* model API installation
 
-Python Model API with model wrappers and pipelines can be installed as a part of OpenVINO&trade; toolkit or from source.
-Installation from source is as follows:
-
-1. Install Python (version 3.6 or higher), [setuptools](https://pypi.org/project/setuptools/):
-
-2. Build the wheel with the following command:
-
-```sh
-python <omz_dir>/demos/common/python/setup.py bdist_wheel
-```
-The built wheel should appear in the dist folder;
-Name example: `openmodelzoo_modelapi-0.0.0-py3-none-any.whl`
-
-3. Install the package in the clean environment with `--force-reinstall` key:
-```sh
-python -m pip install openmodelzoo_modelapi-0.0.0-py3-none-any.whl --force-reinstall
-```
-Alternatively, instead of building the wheel you can use the following command inside  `<omz_dir>/demos/common/python/` directory to build and install the package:
-```sh
-python -m pip install .
-```
-
-When the model API package is installed, you can import it as follows:
-```sh
-python -c "from openvino.model_zoo import model_api"
-```
-
-> **NOTE**: On Linux and macOS, you may need to type `python3` instead of `python`. You may also need to [install pip](https://pip.pypa.io/en/stable/installation/).
-> For example, on Ubuntu execute the following command to get pip installed: `sudo apt install python3-pip`.
+To run Python demo applications, you need to install the Python* Model API package. Refer to [Python* Model API documentation](common/python/openvino/model_zoo/model_api/README.md#installing-python*-model-api-package) to learn about its installation.
 
 ### <a name="build_python_extensions"></a>Build the Native Python\* Extension Modules
 

demos/common/python/openvino/model_zoo/model_api/README.md

@@ -0,0 +1,150 @@
+# Python* Model API package
+
+Model API package is a set of wrapper classes for particular tasks and model architectures, simplifying data preprocess and postprocess as well as routine procedures (model loading, asynchronous execution, etc...)
+An application feeds model class with input data, then the model returns postprocessed output data in user-friendly format.
+
+## Package structure
+
+The Python* Model API consists of 3 libraries:
+* _adapters_ implements a common interface to allow Model API wrappers usage with different executors: OpenVINO, OVMS. See [Model API adapters](#model-api-adapters) section
+* _models_ implements wrappers for each architecture. See [Model API Wrappers](#model-api-wrappers) section
+* _pipelines_ implements pipelines for model inference and manage the synchronous/asynchronous execution. See [Model API Pipelines](#model-api-pipelines) section
+
+### Prerequisites
+
+The package requires
+- Python (version 3.6 or higher)
+- OpenVINO™ toolkit
+
+If you build Python* Model API package from source, you should install the OpenVINO™ toolkit. See the options:
+
+Use installation package for [Intel® Distribution of OpenVINO™ toolkit](https://www.intel.com/content/www/us/en/developer/tools/openvino-toolkit-download.html) or build the open-source version available in the [OpenVINO GitHub repository](https://github.com/openvinotoolkit/openvino) using the [build instructions](https://github.com/openvinotoolkit/openvino/wiki/BuildingCode).
+
+Also, you can install the OpenVINO Python\* package via the command:
+ ```sh
+pip install openvino
+ ```
+
+## Installing Python* Model API package
+
+Use the following command to install Python* Model API from source:
+```sh
+pip install <omz_dir>/demos/common/python
+```
+
+Alternatively, you can generate the package using a wheel. Follow the steps below:
+1. Build the wheel.
+
+```sh
+python <omz_dir>/demos/common/python/setup.py bdist_wheel
+```
+The wheel should appear in the dist folder.
+Name example: `openmodelzoo_modelapi-0.0.0-py3-none-any.whl`
+
+2. Install the package in the clean environment with `--force-reinstall` key.
+```sh
+pip install openmodelzoo_modelapi-0.0.0-py3-none-any.whl --force-reinstall
+```
+
+To verify the package is installed, you might use the following command:
+```sh
+python -c "from openvino.model_zoo import model_api"
+```
+
+## Model API Wrappers
+
+The Python* Model API package provides model wrappers, which implement standardized preprocessing/postprocessing functions per "task type" and incapsulate model-specific logic for usage of different models in a unified manner inside the application.
+
+The wrapper interface is simple and flexible, which gives capabilities for the creation of custom wrappers covering different architectures and use cases.
+
+The following tasks can be solved with wrappers usage:
+
+| Task type                  | Model API wrappers |
+|----------------------------|--------------------|
+| Background Matting         | <ul><li>`VideoBackgroundMatting`</li><li>`ImageMattingWithBackground`</li></ul> |
+| Classification             | <ul><li>`Classification`</li></ul> |
+| Deblurring                 | <ul><li>`Deblurring`</li></ul> |
+| Human Pose Estimation      | <ul><li>`HpeAssociativeEmbedding`</li><li>`OpenPose`</li></ul> |
+| Instance Segmentation      | <ul><li>`MaskRCNNModel`</li><li>`YolactModel`</li></ul> |
+| Monocular Depth Estimation | <ul><li> `MonoDepthModel`</li></ul> |
+| Named Entity Recognition   | <ul><li>`BertNamedEntityRecognition`</li></ul> |
+|  Object Detection          | <ul><li>`CenterNet`</li><li>`DETR`</li><li>`CTPN`</li><li>`FaceBoxes`</li><li>`RetinaFace`</li><li>`RetinaFacePyTorch`</li><li>`SSD`</li><li>`UltraLightweightFaceDetection`</li><li>`YOLO`</li><li>`YoloV3ONNX`</li><li>`YoloV4`</li><li>`YOLOF`</li><li>`YOLOX`</li></ul> |
+| Question Answering         |  <ul><li>`BertQuestionAnswering`</li></ul> |
+| Salient Object Detection   |  <ul><li>`SalientObjectDetectionModel`</li></ul> |
+| Semantic Segmentation      |  <ul><li>`SegmentationModel`</li></ul> |
+
+## Model API Adapters
+
+Model API wrappers are executor-agnostic, meaning it does not implement the specific model inference or model loading, instead it can be used with different executors having the implementation of common interface methods in adapter class respectively.
+
+Currently, `OpenvinoAdapter` and `OVMSAdapter` are supported.
+
+#### OpenVINO Adapter
+
+`OpenvinoAdapter` hides the OpenVINO™ toolkit API, which allows Model API wrappers launching with models represented in Intermediate Representation (IR) format.
+It accepts a path to either `xml` model file or `onnx` model file.
+
+#### OpenVINO Model Server Adapter
+
+`OVMSAdapter` hides the OpenVINO Model Server python client API, which allows Model API wrappers launching with models served by OVMS.
+
+Refer to __[`OVMSAdapter`](adapters/ovms_adapter.md)__ to learn about running demos with OVMS.
+
+For OpenVINO Model Server Adapter employment, you need to install the package with extra module:
+```sh
+pip install <omz_dir>/demos/common/python[ovms]
+```
+
+## Model API Pipelines
+
+Model API Pipelines represent the high-level wrappers upon the input data and accessing model results management.
+They perform the data submission for model inference, verification of inference status, whether the result is ready or not, and results accessing.
+
+The `AsyncPipeline` is available, which handles the asynchronous execution of a single model.
+
+## Ready-to-use Model API solutions
+
+To apply Model API wrappers in custom applications, learn the provided example of common scenario of how to use Python* Model API.
+
+ In the example, the SSD architecture is used to predict bounding boxes on input image `"sample.png"`. The model execution is produced by `OpenvinoAdapter`, therefore we submit the path to the model's `xml` file.
+
+Once the SSD model wrapper instance is created, we get the predictions by the model in one line: `ssd_model(input_data)` - the wrapper performs the preprocess method, synchronous inference on OpenVINO™ toolkit side and postprocess method.
+
+```python
+import cv2
+# import model wrapper class
+from openvino.model_zoo.model_api.models import SSD
+# import inference adapter and helper for runtime setup
+from openvino.model_zoo.model_api.adapters import OpenvinoAdapter, create_core
+
+
+# read input image using opencv
+input_data = cv2.imread("sample.png")
+
+# define the path to mobilenet-ssd model in IR format
+model_path = "public/mobilenet-ssd/FP32/mobilenet-ssd.xml"
+
+# create adapter for OpenVINO™ runtime, pass the model path
+model_adapter = OpenvinoAdapter(create_core(), model_path, device="CPU")
+
+# create model API wrapper for SSD architecture
+# preload=True loads the model on CPU inside the adapter
+ssd_model = SSD(model_adapter, preload=True)
+
+# apply input preprocessing, sync inference, model output postprocessing
+results = ssd_model(input_data)
+```
+
+To study the complex scenarios, refer to [Open Model Zoo Python* demos](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos), where the asynchronous inference is applied.
+
+The list of Open Model Zoo demos with Model API support:
+- [BERT Named Entity Recognition Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/bert_named_entity_recognition_demo/python)
+- [BERT Question Answering Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/bert_question_answering_demo/python)
+- [BERT Question Answering Embedding Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/bert_question_answering_embedding_demo/python)
+- [Classification Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/classification_demo/python)
+- [Image Deblurring Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/deblurring_demo/python)
+- [Human Pose Estimation Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/human_pose_estimation_demo/python)
+- [Instance Segmentation Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/instance_segmentation_demo/python)
+- [MonoDepth Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/monodepth_demo/python)
+- [Object Detection Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/object_detection_demo/python)
+- [Image Segmentation Python* Demo](https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/segmentation_demo/python)

demos/common/python/setup.py

@@ -30,6 +30,9 @@
 with open(SETUP_DIR / 'requirements.txt') as f:
     required = f.read().splitlines()
 
+with open(SETUP_DIR / 'requirements_ovms.txt') as f:
+    ovms_required = f.read().splitlines()
+
 packages = find_packages(str(SETUP_DIR))
 package_dir = {'openvino': str(SETUP_DIR / 'openvino')}
 
@@ -47,4 +50,5 @@
     packages=packages,
     package_dir=package_dir,
     install_requires=required,
+    extras_require={'ovms': ovms_required}
 )

demos/object_detection_demo/python/README.md

@@ -39,6 +39,12 @@ Async API operates with a notion of the "Infer Request" that encapsulates the in
 
 > **NOTE**: By default, Open Model Zoo demos expect input with BGR channels order. If you trained your model to work with RGB order, you need to manually rearrange the default channels order in the demo application or reconvert your model using the Model Optimizer tool with the `--reverse_input_channels` argument specified. For more information about the argument, refer to **When to Reverse Input Channels** section of [Converting a Model Using General Conversion Parameters](https://docs.openvino.ai/latest/openvino_docs_MO_DG_prepare_model_convert_model_Converting_Model.html#general-conversion-parameters).
 
+## Model API
+
+The demo utilizes model wrappers, adapters and pipelines from [Python* Model API](../../common/python/openvino/model_zoo/model_api/README.md).
+
+The generalized interface of wrappers with its unified results representation provides the support of multiple different object detection model topologies in one demo.
+
 ## Preparing to Run
 
 For demo input image or video files, refer to the section **Media Files Available for Demos** in the [Open Model Zoo Demos Overview](../../README.md).