Merge branch 'openvinotoolkit:master' into master

Author: ieliz

CONTRIBUTING.md

@@ -46,7 +46,7 @@ File | Destination
 configuration file | `models/public/<model_name>/model.yml`
 documentation file | `models/public/<model_name>/README.md`
 validation configuration file|`models/public/<model_name>/accuracy-check.yml`
-demo|`demos/<demo_name>`<br>or<br>`demos/python_demos/<demo_name>`
+demo|`demos/<demo_name>/<language>/`
 
 ### Tests
 
@@ -185,22 +185,19 @@ Deep Learning Inference Engine (IE) supports models in the Intermediate Represen
 
 ## Demo
 
-A demo shows the main idea of how to infer a model using IE. If your model solves one of the tasks supported by the Open Model Zoo, try to find an appropriate option from [demos](demos/README.md) or [samples](https://docs.openvino.ai/latest/_docs_IE_DG_Samples_Overview.html). Otherwise, you must provide your own demo (C++ or Python).
+A demo shows the main idea of how to infer a model using OpenVINO™. If your model solves one of the tasks supported by the Open Model Zoo, try to find an appropriate option from [demos](demos/README.md) or [samples](https://docs.openvino.ai/latest/_docs_IE_DG_Samples_Overview.html). Otherwise, you must provide your own demo (C++ or Python).
 
 The demo's name should end with `_demo` suffix to follow the convention of the project.
 
-Demos are required to support the following keys:
+Demos are required to support the following args:
 
- -  `-i "<input>"`: Required. An input to process. The input can usually be a single image, a folder of images or anything that OpenCV's `VideoCapture` can process.
- -  `-m "<path>"`: Required. Path to an .xml file with a trained model. If the demo uses several models at the same time, use other keys prefixed with `-m_`.
- -  `-d "<device>"`: Optional. Specifies a target device to infer on. CPU, GPU, HDDL or MYRIAD is acceptable. Default must be CPU. If the demo uses several models at the same time, use keys prefixed with `d_` (just like keys `m_*` above) to specify device for each model.
- -  `-no_show`: Optional. Do not visualize inference results.
+* `-h, --help`        show this help message and exit
+* `-m <MODEL FILE>`   path to an .xml file with a trained model. If the demo uses several models an extended syntax can be used, like `--mdet`
+* `-i <INPUT>`        input to process. For vision tasks the input might be a path to single image or video file, a path to folder of images, or numeric camera id. For vision tasks the default value must be `0`. For speech/audio tasks input is path to WAV file. For NLP tasks input might be a path to text file or just quoted sentence of text.
+* `-d <DEVICE>`       specify a device to infer on (the list of available devices is shown below). Default is CPU
+* `-o <FILE PATTERN>` pattern for output file(s) to save
 
-> **TIP**: For Python, it is preferable to use `--` instead of `-` for long keys.
-
-You can also add any other necessary parameters.
-
-Add `README.md` file, which describes demo usage. Update [demos' README.md](demos/README.md) adding your demo to the list.
+Add `README.md` file, which describes demo usage. Update [demos' README.md](demos/README.md) by adding your demo to the list.
 
 ## Accuracy Validation
 

ci/dependencies.yml

@@ -1,4 +1,6 @@
-openvino_linux: '2022.1.0.518'
-openvino_windows: '2022.1.0.518'
-wheel_linux: '2022.1.0.dev20220117-6152'
-wheel_windows: '2022.1.0.dev20220117-6152'
+opencv_linux: '2022.1.0.dev20220131'
+opencv_windows: '2022.1.0.dev20220131'
+openvino_linux: '2022.1.0.dev20220131'
+openvino_windows: '2022.1.0.dev20220131'
+wheel_linux: '2022.1.0.dev20220131-6386'
+wheel_windows: '2022.1.0.dev20220131-6386'

ci/prepare-documentation.py

@@ -377,6 +377,11 @@ def main():
 
     add_demos_pages(output_root, navindex_element)
 
+    ovms_adapter_element = add_page(output_root, navindex_element, id='omz_model_api_ovms_adapter',
+        path='demos/common/python/openvino/model_zoo/model_api/adapters/ovms_adapter.md',
+        title='OMZ Model API OVMS adapter')
+    ovms_adapter_element.attrib[XML_ID_ATTRIBUTE] = 'omz_model_api_ovms_adapter'
+
     for md_path in all_md_paths:
         if md_path not in documentation_md_paths:
             raise RuntimeError(f'{all_md_paths[md_path]}: '

ci/requirements-conversion.txt

@@ -87,7 +87,7 @@ onnx==1.10.1
     #   -r tools/model_tools/requirements-pytorch.in
 opt-einsum==3.3.0
     # via tensorflow
-paddlepaddle==2.1.3
+paddlepaddle==2.2.0
     # via
     #   -r tools/model_tools/requirements-paddle.in
 pillow==8.3.2

ci/requirements-demos.txt

@@ -154,6 +154,7 @@ xmltodict==0.12.0
     # via motmetrics
 zipp==3.5.0
     # via importlib-metadata
+inflect==5.3.0
 
 # The following packages are considered to be unsafe in a requirements file:
 # setuptools

ci/requirements-downloader.txt

@@ -8,8 +8,6 @@ idna==3.2
     # via requests
 openvino-telemetry==2022.1.0
     # via -r tools/model_tools/requirements.in
-pyrx==0.3.0
-    # via -r tools/model_tools/requirements.in
 pyyaml==5.4.1
     # via -r tools/model_tools/requirements.in
 requests==2.26.0

data/dataset_definitions.yml

@@ -12,11 +12,11 @@ datasets:
   - name: ms_coco_mask_rcnn_short_80_classes
     annotation_conversion:
       converter: mscoco_mask_rcnn
-      annotation_file: annotations/instances_val2017_short.json
+      annotation_file: annotations/instances_val2017.json
       has_background: True
       sort_annotations: True
-    annotation: mscoco_mask_rcnn_short_80.pickle
-    dataset_meta: mscoco_mask_rcnn_short_80.json
+    annotation: mscoco_mask_rcnn_80.pickle
+    dataset_meta: mscoco_mask_rcnn_80.json
     data_source: val2017
 
   - name: ms_coco_mask_rcnn_short_80_classes_without_background
@@ -25,19 +25,19 @@ datasets:
       annotation_file: annotations/instances_val2017.json
       has_background: False
       sort_annotations: True
-    annotation: mscoco_mask_rcnn_short_80_without_bkgr.pickle
-    dataset_meta: mscoco_mask_rcnn_short_80_without_bkgr.json
+    annotation: mscoco_mask_rcnn_80_without_bkgr.pickle
+    dataset_meta: mscoco_mask_rcnn_80_without_bkgr.json
     data_source: val2017
 
   - name: ms_coco_mask_rcnn_short_91_classes
     annotation_conversion:
       converter: mscoco_mask_rcnn
-      annotation_file: annotations/instances_val2017_short.json
+      annotation_file: annotations/instances_val2017.json
       has_background: True
       sort_annotations: True
       use_full_label_map: True
-    annotation: mscoco_mask_rcnn_short_91.pickle
-    dataset_meta: mscoco_mask_rcnn_short_91.json
+    annotation: mscoco_mask_rcnn_91.pickle
+    dataset_meta: mscoco_mask_rcnn_91.json
     data_source: val2017
     preprocessing:
       - type: resize
@@ -1437,3 +1437,12 @@ datasets:
       data_dir: LIVE_1
       input_suffix: .bmp
       target_suffix: .bmp
+
+  - name: online_mstcn_plus_dataset
+    data_source: annotation
+    reader: numpy_dict_reader
+    annotation_conversion:
+      converter: multi_feature_regression
+      data_dir: annotation
+      input_suffix: in
+      reference_suffix: out

demos/3d_segmentation_demo/python/3d_segmentation_demo.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
- Copyright (c) 2019 Intel Corporation
+ Copyright (c) 2019-2022 Intel Corporation
 
  Licensed under the Apache License, Version 2.0 (the "License");
  you may not use this file except in compliance with the License.
@@ -28,7 +28,7 @@
 from fnmatch import fnmatch
 from scipy.ndimage import interpolation
 
-from openvino.inference_engine import IECore, get_version
+from openvino.runtime import Core, get_version, PartialShape
 
 
 log.basicConfig(format='[ %(levelname)s ] %(message)s', level=log.DEBUG, stream=sys.stdout)
@@ -248,48 +248,41 @@ def main():
     # --------------------------------- 1. Load Plugin for inference engine ---------------------------------
     log.info('OpenVINO Inference Engine')
     log.info('\tbuild: {}'.format(get_version()))
-    ie = IECore()
+    core = Core()
 
     if 'CPU' in args.target_device:
         if args.path_to_extension:
-            ie.add_extension(args.path_to_extension, "CPU")
+            core.add_extension(args.path_to_extension, "CPU")
         if args.number_threads is not None:
-            ie.set_config({'CPU_THREADS_NUM': str(args.number_threads)}, "CPU")
+            core.set_config({'CPU_THREADS_NUM': str(args.number_threads)}, "CPU")
     elif 'GPU' in args.target_device:
         if args.path_to_cldnn_config:
-            ie.set_config({'CONFIG_FILE': args.path_to_cldnn_config}, "GPU")
+            core.set_config({'CONFIG_FILE': args.path_to_cldnn_config}, "GPU")
     else:
         raise AttributeError("Device {} do not support of 3D convolution. "
                              "Please use CPU, GPU or HETERO:*CPU*, HETERO:*GPU*")
 
     # --------------------- 2. Read IR Generated by ModelOptimizer (.xml and .bin files) ---------------------
     log.info('Reading model {}'.format(args.path_to_model))
-    ie_network = ie.read_network(args.path_to_model, os.path.splitext(args.path_to_model)[0] + '.bin')
+    model = core.read_model(args.path_to_model)
 
-    input_info = ie_network.input_info
-    if len(input_info) == 0:
-        raise AttributeError('No inputs info is provided')
-    elif len(input_info) != 1:
-        raise AttributeError("only one input layer network is supported")
-
-    input_name = next(iter(input_info))
-    out_name = next(iter(ie_network.outputs))
+    if len(model.inputs) != 1:
+        raise RuntimeError("only 1 input layer model is supported")
 
+    input_tensor_name = model.inputs[0].get_any_name()
     if args.shape:
-        log.debug("Reshape model from {} to {}".format(input_info[input_name].input_data.shape, args.shape))
-        ie_network.reshape({input_name: args.shape})
-        input_info = ie_network.input_info
+        log.debug("Reshape model from {} to {}".format(model.inputs[0].shape, args.shape))
+        model.reshape({input_tensor_name: PartialShape(args.shape)})
 
-    if len(input_info[input_name].input_data.shape) != 5:
-        raise AttributeError("Incorrect shape {} for 3d convolution network".format(args.shape))
+    if len(model.inputs[0].shape) != 5:
+        raise RuntimeError("Incorrect shape {} for 3d convolution network".format(args.shape))
 
-    n, c, d, h, w = input_info[input_name].input_data.shape
-    ie_network.batch_size = n
+    n, c, d, h, w = model.inputs[0].shape
 
     # ------------------------------------ 3. Loading model to the plugin -------------------------------------
-    executable_network = ie.load_network(network=ie_network, device_name=args.target_device)
+    compiled_model = core.compile_model(model, args.target_device)
+    infer_request = compiled_model.create_infer_request()
     log.info('The model {} is loaded to {}'.format(args.path_to_model, args.target_device))
-    del ie_network
 
     # --------------------------------------- 4. Preparing input data -----------------------------------------
     start_time = perf_counter()
@@ -319,12 +312,10 @@ def main():
         original_data = data_crop
         original_size = original_data.shape[-3:]
 
-    test_im = {input_name: data_crop}
-
     # ---------------------------------------------- 5. Do inference --------------------------------------------
-    res = executable_network.infer(test_im)
+    result = infer_request.infer({input_tensor_name: data_crop})
     # ---------------------------- 6. Processing of the received inference results ------------------------------
-    result = res[out_name]
+    result = next(iter(result.values()))
     batch, channels, out_d, out_h, out_w = result.shape
 
     list_img = []
@@ -397,4 +388,4 @@ def main():
             log.debug("Result nifti file was saved to {}".format(nii_filename))
 
 if __name__ == "__main__":
-    main()
+    sys.exit(main() or 0)

demos/3d_segmentation_demo/python/requirements.txt

@@ -12,6 +12,7 @@ The Open Model Zoo includes the following demos:
 - [3D Segmentation Python\* Demo](./3d_segmentation_demo/python/README.md) - Segmentation demo segments 3D images using 3D convolutional networks.
 - [Action Recognition Python\* Demo](./action_recognition_demo/python/README.md) - Demo application for Action Recognition algorithm, which classifies actions that are being performed on input video.
 - [Background Subtraction Python\* Demo](./background_subtraction_demo/python/README.md) - Background subtraction using instance segmentation based models.
+- [Background Subtraction C++ G-API\* Demo](./background_subtraction_demo/cpp_gapi/README.md) - Background subtraction G-API version.
 - [BERT Named Entity Recognition Python\* Demo](./bert_named_entity_recognition_demo/python/README.md) - NER Demo application that uses a CONLL2003-tuned BERT model for inference.
 - [BERT Question Answering Python\* Demo](./bert_question_answering_demo/python/README.md)
 - [BERT Question Answering Embedding Python\* Demo](./bert_question_answering_embedding_demo/python/README.md) - The demo demonstrates how to run BERT based models for question answering task.